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
INSIGHT 2 was a phase 2, non-randomized, open-label study in non-small cell lung cancer evaluating tepotinib and osimertinib in a parallel two-arm design. The registry reports an enrollment of 140 participants and identifies the study as completed.
| Feature | INSIGHT 2 |
|---|---|
| Phase | Phase 2 |
| Condition | Non-small Cell Lung Cancer |
| Design | Non-randomized, parallel, two-arm |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 140 |
| Interventions | Tepotinib; Osimertinib |
| Status | Completed |
| Lead sponsor | EMD Serono Research & Development Institute, Inc. |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT03940703 |
2. Clinical Question
The central clinical question was whether the study regimen involving tepotinib plus osimertinib could produce measurable antitumor activity in participants with osimertinib-relapsed MET-amplified non-small cell lung cancer, while also characterizing dose-limiting toxicities during the first cycle.
Population
Participants with osimertinib-relapsed MET-amplified non-small cell lung cancer.
Intervention
Tepotinib plus osimertinib, identified in the registry as the combined therapy.
Comparator / second arm
Tepotinib mono-therapy is identified as the second arm in the registry safety reporting.
Primary questions
How many participants experienced dose-limiting toxicities during Cycle 1, and what proportion experienced an objective response according to RECIST Version 1.1?
3. Trial Design
Tepotinib + Osimertinib
- Combined tepotinib and osimertinib therapy
- Primary DLT assessment through Day 21 of Cycle 1
- Objective response assessed according to RECIST Version 1.1
- Objective response evaluated by an Independent Review Committee in participants with centrally determined MET amplification by FISH
Tepotinib Mono-therapy
- Tepotinib mono-therapy
- Included as the second arm in the registry safety reporting
- Serious adverse events are reported for this arm as 2/12
4. Trial Timing and Registry Status
Study start
The registry lists September 19, 2019 as the study start date.
Primary completion
The registry lists May 11, 2023 as the primary completion date.
Current registry status
The study is listed as completed.
5. Primary Endpoints
| Endpoint | Registry definition / assessment | Time frame |
|---|---|---|
| Combined Therapy: Number of Participants Experiencing Dose-Limiting Toxicities (DLTs) | DLTs are defined as specified hematologic and non-hematologic toxicities judged by the Investigator and/or Sponsor to be not attributable to the disease or disease-related processes under investigation. The definition includes Grade 4 neutropenia for more than 7 days; Grade ≥3 febrile neutropenia; Grade 4 thrombocytopenia or Grade 3 thrombocytopenia with non-traumatic bleeding; Grade ≥3 nausea/vomiting and/or diarrhea not improved within 72 hours despite adequate and optimal treatment; and other Grade ≥3 non-hematological adverse events, with specified exceptions. | Up to Day 21 of Cycle 1; each Cycle is 21 days |
| Combined Therapy: Objective Response According to RECIST Version 1.1 | Objective response is defined as the percentage of participants with either a confirmed complete response or partial response from first administration of study treatment to first observation of progressive disease. Assessment is by Independent Review Committee in participants with MET amplification determined centrally by FISH. | From first administration of study treatment up to data cutoff; approximately assessed up to 42 months |
Dose-Limiting Toxicity
The DLT endpoint is a binary safety endpoint: each participant is classified according to whether a protocol-defined dose-limiting toxicity occurred within the specified Cycle 1 window. The endpoint therefore focuses on early treatment tolerability rather than cumulative toxicity over the entire study.
Objective Response
The objective-response endpoint is also binary at the participant level. A participant is considered to have an objective response when a confirmed complete response or partial response is observed according to the registry definition. The denominator is important because the registry describes this as a percentage of participants rather than a continuous tumor-measurement outcome.
6. Planned Analysis
The registry reports results as posted for the study, but no formal statistical analyses are posted in the ClinicalTrials.gov record. The registry provides the endpoint definitions and the assessment framework rather than treatment-effect estimates, confidence intervals, or hypothesis-test results.
| Endpoint | Statistical structure | Typical analysis |
|---|---|---|
| DLTs through Day 21 of Cycle 1 | Binary participant-level safety outcome | Descriptive incidence: number and percentage of participants with a DLT. A binomial confidence interval can quantify precision around the observed proportion. |
| Objective response | Binary tumor-response outcome | Objective response rate is ordinarily summarized as the number and percentage of responders. A binomial confidence interval can quantify uncertainty around the response proportion. |
For a binary endpoint, the basic estimand is a proportion:
Here, x is the number of participants meeting the endpoint definition and n is the relevant analysis denominator. The proportion is descriptive unless a prespecified comparison or hypothesis test is defined.
7. Dose-Limiting Toxicity: Statistical Interpretation
The first primary endpoint evaluates whether participants receiving the combined therapy experience predefined DLTs during the first 21 days of treatment. This is a deliberately narrow time window: it is designed to identify early toxicities that meet the protocol's severity and duration criteria.
What is measured
Whether a participant experiences a toxicity satisfying the protocol's DLT definition through Day 21 of Cycle 1.
Why the window matters
A toxicity occurring outside the specified window is not automatically a DLT for this endpoint, even though it may still be clinically relevant safety information.
Why attribution matters
The registry definition excludes toxicities judged by the Investigator and/or Sponsor to be attributable to the disease or disease-related processes under investigation.
What a proportion means
The DLT proportion describes the frequency of protocol-defined early toxicity in the relevant treatment population; it does not summarize all adverse events over the study.
DLT definition in detail
- Grade 4 neutropenia for more than 7 days.
- Grade ≥3 febrile neutropenia.
- Grade 4 thrombocytopenia or Grade 3 thrombocytopenia with non-traumatic bleeding.
- Grade ≥3 nausea/vomiting and/or diarrhea that has not improved within 72 hours despite adequate and optimal treatment.
- Any other Grade ≥3 non-hematological adverse event, except the specified exclusions in the registry definition.
8. Objective Response: Statistical Interpretation
The second primary endpoint evaluates objective response according to RECIST Version 1.1. The registry defines response as the percentage of participants with a confirmed complete response or partial response from first administration of study treatment to first observation of progressive disease.
| Response category | Registry definition |
|---|---|
| Complete response (CR) | Disappearance of all evidence of target and non-target lesions. |
| Partial response (PR) | At least 30% reduction from baseline in the sum of the longest diameter of all lesions. |
| Progressive disease (PD) | At least a 20% increase in the sum of the longest diameter, taking as reference the smallest sum recorded from baseline, or appearance of one or more new lesions. |
| Objective response | Confirmed CR or PR during the specified assessment period. |
Because objective response is a categorical endpoint, its analysis differs fundamentally from analysis of a continuous tumor-size measurement. The key quantity is the proportion of participants satisfying the response definition. A confidence interval around that proportion describes statistical precision, whereas the proportion itself describes the observed frequency of response.
The exact denominator depends on the prespecified analysis population. The registry excerpt identifies the endpoint and assessment framework but does not provide a posted statistical-analysis table defining an observed response estimate and its confidence interval.
9. Serious Adverse Events
The registry reports serious adverse events by arm using affected participants over participants at risk. These data provide an important safety perspective that is distinct from the Cycle 1 DLT endpoint.
| Arm | Affected / at risk | Interpretation |
|---|---|---|
| Tepotinib and Osimertinib | 49/128 | 49 participants affected among 128 at risk for the reported serious-adverse-event measure. |
| Tepotinib Mono-therapy | 2/12 | 2 participants affected among 12 at risk for the reported serious-adverse-event measure. |
Reported serious adverse-event counts
Tepotinib and Osimertinib vs Tepotinib Mono-therapy
The denominators are particularly important here. The registry reports 128 participants at risk for the combined-therapy safety measure and 12 at risk for tepotinib mono-therapy. These are the denominators posted on ClinicalTrials.gov for the serious-adverse-event reporting and should not be replaced with the overall enrollment simply because enrollment was 140.
10. Statistical Methods Explained
Why is the study's non-randomized design statistically important?
Randomization creates a mechanism for balancing measured and unmeasured prognostic factors across treatment groups in expectation. INSIGHT 2 is classified as non-randomized, so that protection is absent. A difference in an endpoint between the two arms can therefore reflect treatment, patient selection, disease characteristics, or other differences between the groups.
Why is objective response analyzed as a proportion?
RECIST response places each participant into clinically defined categories. The primary quantity is therefore the fraction of participants achieving a confirmed CR or PR. The proportion is straightforward to communicate, while a binomial confidence interval adds information about its precision.
Why does the RECIST definition use a 30% reduction for partial response?
The registry defines PR as at least a 30% reduction from baseline in the sum of the longest diameter of all lesions. This creates a prespecified threshold separating a protocol-defined response from smaller changes in tumor measurements. The threshold is therefore part of the endpoint definition, not a cutoff selected after seeing the results.
Why is progressive disease defined relative to the smallest recorded sum?
The registry defines PD using at least a 20% increase in the sum of the longest diameter, taking the smallest sum recorded from baseline as the reference, or the appearance of one or more new lesions. Using the smallest prior measurement helps identify meaningful subsequent tumor growth rather than treating the original baseline measurement as the only reference throughout follow-up.
Why is a confidence interval useful for a response rate?
A response rate calculated from a finite number of participants is subject to sampling uncertainty. A confidence interval gives a range of values compatible with the observed data under the specified statistical framework. It does not mean that individual patients have probabilities spread uniformly across that interval, nor does it establish causality.
Why can a non-randomized comparison not be interpreted like a randomized treatment effect?
In a randomized trial, the treatment assignment mechanism supports a causal comparison under the trial assumptions. In a non-randomized study, the treatment groups may differ systematically before treatment. Statistical adjustment can address measured differences under additional assumptions, but it cannot automatically eliminate unmeasured confounding.
11. Statistical Methodology
Descriptive analysis of binary endpoints
Both primary endpoints have a binary participant-level structure. The natural descriptive analysis is the number and percentage of participants meeting the endpoint definition.
The count x records participants with the endpoint and n is the appropriate analysis denominator. For statistical inference, exact or other binomial confidence intervals may be used depending on the prespecified analysis plan.
Confidence intervals for proportions
For an objective response rate or DLT rate, the width of a confidence interval depends strongly on the number of participants contributing to the denominator. Smaller denominators generally produce less precise estimates. This is particularly relevant when an arm has only 12 participants at risk for a reported safety measure.
RECIST response assessment
RECIST Version 1.1 converts longitudinal radiologic measurements into standardized categories such as complete response, partial response, stable disease, and progressive disease. For this trial's primary response endpoint, the registry specifically identifies confirmed CR or PR as the basis of objective response and specifies assessment by an Independent Review Committee in participants with centrally determined MET amplification by FISH.
Independent Review Committee
An Independent Review Committee provides a review framework separate from the treating investigator's routine assessment. For this endpoint, the registry identifies the committee as the assessor of objective response according to RECIST Version 1.1.
Central MET amplification determination
The objective-response endpoint is specifically defined in participants with MET amplification determined centrally by fluorescence in situ hybridization, or FISH. This creates an important distinction between the broader study population and the population explicitly described in the response endpoint definition.
12. Analysis Population and Denominator Issues
One of the most important statistical details in a registry-based trial summary is identifying the denominator attached to each endpoint. INSIGHT 2 has an overall enrollment of 140, but the serious-adverse-event reporting uses arm-specific at-risk denominators of 128 and 12.
| Quantity | Reported value | Why it matters |
|---|---|---|
| Total enrollment | 140 | Describes the total number enrolled in the study. |
| Serious adverse events, combined therapy | 49/128 | Uses 128 as the reported at-risk denominator. |
| Serious adverse events, tepotinib mono-therapy | 2/12 | Uses 12 as the reported at-risk denominator. |
| Objective response population | Not numerically reported in the registry excerpt | The endpoint definition specifies the MET-amplified population assessed by Independent Review Committee, but no numerical response estimate is posted in the statistical-analysis section. |
Using the overall enrollment as the denominator for every endpoint would therefore be statistically inappropriate. Each endpoint must be interpreted using the population to which that endpoint applies.
13. What a Missing Formal Analysis Means for Interpretation
The ClinicalTrials.gov record identifies results as posted, but it does not contain posted statistical analyses for the two primary endpoints in the record represented here. Consequently, the registry supports interpretation of what was measured and how it was defined, but it does not supply a numerical DLT estimate, objective response estimate, confidence interval, or p-value for those endpoints.
What can be established
The primary endpoints are explicitly defined, including their time frames, RECIST criteria, Independent Review Committee assessment, and central FISH determination for MET amplification.
What requires caution
A study being marked "Results posted: Yes" does not mean that every endpoint has a posted formal statistical analysis with an effect estimate and inferential statistics.
What the safety data show
The registry reports serious adverse-event counts using arm-specific at-risk denominators: 49/128 and 2/12.
What the endpoint design implies
DLT and objective response are naturally summarized as proportions, but the inferential interpretation depends on the prespecified analysis population and statistical procedure.
14. Why the Non-Randomized Design Changes the Statistical Story
INSIGHT 2 differs fundamentally from a randomized controlled trial because allocation is explicitly classified as non-randomized. This affects how treatment-group comparisons should be understood.
| Feature | Statistical consequence |
|---|---|
| No randomization | Baseline differences between treatment groups can be associated with outcome differences. |
| Two parallel arms | Each arm provides a contemporaneous treatment group, but the design does not guarantee exchangeability between arms. |
| No masking | Knowledge of treatment assignment can potentially affect clinical management or assessment, although an Independent Review Committee is identified for the objective-response endpoint. |
| Phase 2 treatment study | The design emphasizes characterization of treatment activity and safety rather than the same confirmatory causal framework used in a randomized phase 3 comparison. |
This distinction is especially important for objective response. Even if one arm has a higher observed response proportion, that observation alone cannot separate a treatment effect from differences in the participants who received each intervention.
15. Limitations
- Non-randomized allocation: treatment groups are not created through random assignment, so between-arm differences cannot automatically be interpreted causally.
- No masking: the registry classifies the study as unmasked. Awareness of treatment assignment can be relevant to some aspects of clinical assessment and management.
- Limited posted inference: the registry identifies the results as posted but does not provide formal statistical analyses for the primary endpoints in the record represented here.
- Endpoint-specific denominators: the safety results use arm-specific at-risk populations of 128 and 12, rather than the overall enrollment of 140.
- Early DLT window: the DLT endpoint is restricted to up to Day 21 of Cycle 1 and therefore does not represent all toxicity occurring throughout treatment.
- Response population: the objective-response endpoint specifically concerns participants with MET amplification determined centrally by FISH, so the endpoint should not automatically be generalized to every enrolled participant without confirming the analysis population.
- Proportion precision: binary endpoint estimates are sensitive to sample size. Smaller denominators produce greater statistical uncertainty.
- RECIST categorization: objective response depends on prespecified radiologic thresholds and confirmation rules, so it is not equivalent to average tumor shrinkage across all participants.
16. Why This Trial Matters Statistically
INSIGHT 2 is a useful teaching case because it illustrates how the statistical interpretation of a phase 2 oncology study depends not only on the endpoint definition, but also on allocation, assessment, and denominator structure.
| Concept | How it appears in INSIGHT 2 |
|---|---|
| Non-randomized design | Participants were not randomly allocated to treatment. |
| Parallel design | The registry identifies two parallel arms. |
| Binary safety endpoint | DLT is defined as whether a participant experiences a protocol-defined toxicity during Cycle 1. |
| Binary efficacy endpoint | Objective response is defined by confirmed CR or PR according to RECIST Version 1.1. |
| Central biomarker determination | MET amplification is determined centrally by FISH for the objective-response endpoint. |
| Independent review | Objective response is assessed by an Independent Review Committee. |
| Time-windowed endpoint | DLT is assessed through Day 21 of Cycle 1. |
| Longer response window | Objective response is assessed from first treatment administration to data cutoff, approximately up to 42 months. |
| Denominator discipline | Serious adverse-event reporting uses 128 and 12 participants at risk for the two reported arms. |
| Phase 2 evidence | The study provides a framework for evaluating treatment activity and safety in the specified population. |
17. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The primary endpoints are binary participant-level outcomes with prespecified definitions. Their natural summaries are proportions, accompanied when appropriate by confidence intervals. Because the study is non-randomized, between-arm comparisons require stronger assumptions than a randomized comparison.
Clinical interpretation
The study addresses both early treatment tolerability and objective tumor response. DLT captures a specific early safety window, while RECIST objective response captures a defined form of tumor response over the broader assessment period.
18. Statistical Concepts in This Trial
Several core statistical ideas can be illustrated directly from the design and endpoint definitions of INSIGHT 2.
Binary endpoints
DLT and objective response reduce participant-level observations to clearly defined yes/no outcomes for the primary endpoint analysis.
Proportion estimation
Both primary endpoints can be summarized as the proportion of participants meeting the relevant endpoint definition.
Confidence intervals
Confidence intervals quantify uncertainty around a sample proportion and should be interpreted separately from the point estimate.
Non-randomized comparison
Observed differences between arms can reflect treatment and underlying differences between the populations receiving each intervention.
RECIST
Standardized radiologic criteria convert changes in lesion measurements into response and progression categories.
Analysis populations
The correct denominator is endpoint-specific; enrollment of 140 does not imply that every endpoint uses 140 participants.
19. Sources
- ClinicalTrials.gov: NCT03940703 — INSIGHT 2.
Continue through the Clinical Biostats knowledge graph
Explore statistical concepts, clinical trial methods, and analysis workflows connected to modern clinical research.
20. Record Summary
INSIGHT 2 is a completed phase 2, non-randomized, open-label, parallel study of tepotinib and osimertinib in osimertinib-relapsed MET-amplified non-small cell lung cancer. Its two primary endpoints focus on different statistical questions: dose-limiting toxicity through Day 21 of Cycle 1 and objective response according to RECIST Version 1.1 over the period from first treatment administration to data cutoff, approximately up to 42 months.
The statistical structure is therefore centered on binary endpoint analysis, proportion estimation, RECIST classification, and careful attention to the relevant analysis denominator. The non-randomized design is equally important: observed differences between arms should not be interpreted in the same causal framework as differences produced by random treatment assignment. The registry reports serious adverse events as 49/128 for tepotinib and osimertinib and 2/12 for tepotinib mono-therapy, while the primary endpoint definitions provide the framework for interpreting DLT and objective response.