This page separates registry-reported trial information from statistical interpretation. The ClinicalTrials.gov record contains registered primary endpoints and posted outcome measures, but no formal statistical analyses in the statistical-analysis field.
1. Trial at a Glance
SUMMIT was a phase 2, non-randomized, parallel study evaluating neratinib-based treatment strategies in participants with solid tumors harboring somatic HER2 or EGFR exon 18 mutations. The ClinicalTrials.gov record reports an enrollment of 582 participants, six arms, three registered primary endpoints, and twelve posted outcome measures.
| Feature | SUMMIT |
|---|---|
| Trial name | SUMMIT |
| ClinicalTrials.gov identifier | NCT01953926 |
| Brief title | Basket Study of Neratinib in Participants With Solid Tumors Harboring Somatic HER2 or EGFR Exon 18 Mutations |
| Phase | Phase 2 |
| Status | Terminated |
| Start | 2013-09-30 |
| Primary completion | 2023-01-02 |
| Lead sponsor | Puma Biotechnology, Inc. |
| Sponsor type | Industry |
| Allocation | Non-randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Conditions | Solid Tumors Harboring Somatic HER2 or EGFR Exon 18 Mutations |
| Interventions | Neratinib; Fulvestrant; Trastuzumab; Paclitaxel |
2. Clinical Question
The SUMMIT registry describes a basket study evaluating treatment in participants with solid tumors harboring somatic HER2 or EGFR exon 18 mutations. Because allocation was non-randomized and the study used multiple arms, the statistical question is not a single randomized comparison between one experimental group and one control group. Instead, the principal registered efficacy questions concern whether participants in specified disease or mutation cohorts achieve a confirmed objective response under the applicable response-assessment framework.
Population
Participants with solid tumors harboring somatic HER2 or EGFR exon 18 mutations.
Interventions
The ClinicalTrials.gov record identifies neratinib, fulvestrant, trastuzumab, and paclitaxel as study interventions.
Design
Phase 2, non-randomized, parallel, unmasked study with a treatment purpose.
Primary statistical question
What proportion of participants achieve complete or partial response under the registry-defined objective-response endpoint for the relevant cohort?
3. Trial Design
4. Registered Primary Endpoints
The ClinicalTrials.gov record identifies three primary endpoints. All three are binary endpoints, meaning the primary outcome is classified as response or non-response according to the endpoint-specific response definition.
| Endpoint | Time frame | Endpoint type | Assessment |
|---|---|---|---|
| Confirmed Objective Response Rate (ORR) by Independent Central Review (Breast Cancer With Prior CDK46i Cohort) | From enrollment date to first confirmed Complete or Partial Response, whichever came earlier, assessed up to 58 months | Binary | Percentage of participants confirmed by independent central review to have achieved CR or PR according to RECIST v1.1. The registry definition identifies HR+, HER2 negative metastatic breast cancer cohorts and target-lesion assessment by MRI or CT. |
| Confirmed Objective Response Rate (ORR) by Investigator Review (Cervical Cancer Cohort) | From enrollment date to first confirmed Complete or Partial Response, whichever came earlier, assessed up to 58 months | Binary | Percentage of participants confirmed by investigator review to have achieved CR or PR according to RECIST v1.1 in the cervical cancer cohort. The registry definition identifies target-lesion assessment by MRI or CT. |
| Objective Response Rate (ORR) at First Assessment by Investigator Review (All Other Cohorts) | From first treatment date to first Complete or Partial Response, whichever came earlier, assessed up to 8 or 9 weeks | Binary | Percentage of participants who achieve CR or PR according to RECIST v1.1, or other defined response criteria, at the first scheduled tumor assessment. The registry definition identifies RECIST, if assessed, or PERCIST. |
Why the endpoint definitions differ
The three primary endpoints are not interchangeable. Two use confirmation and extend assessment up to 58 months, while the third evaluates response at the first scheduled tumor assessment within 8 or 9 weeks. The first also uses independent central review, whereas the other two use investigator review.
Those differences affect interpretation. A confirmed ORR evaluated over a long observation window is not statistically identical to a first-assessment ORR. Similarly, independent central review and investigator review represent different measurement processes. Comparing their numerical values as though they were produced by one common endpoint would obscure the design of the study.
5. Statistical Methodology
Binary objective-response endpoints
Each registered primary endpoint is binary: a participant either achieves the specified complete or partial response or does not. The basic estimand is therefore a proportion:
For a binary endpoint, uncertainty can be quantified with a binomial confidence interval. The registry extract does not provide formal statistical-analysis results for the three primary endpoints.
Response assessment
The registry definitions use RECIST v1.1 for the specified cohorts, with target-lesion assessment by MRI or CT where described. For the all-other-cohorts endpoint, the registry permits RECIST v1.1 or other defined response criteria and identifies RECIST, if assessed, or PERCIST.
Independent central review versus investigator review
Independent central review introduces a separate assessment process intended to reduce the influence of knowledge and local interpretation on response classification. Investigator review reflects assessment by the study investigators. The distinction is statistically relevant because the outcome itself depends on the measurement process used to classify response.
Non-randomized treatment allocation
Because the registry-reported design is non-randomized, the principal statistical challenge is different from that of a conventional randomized comparison. Within-cohort ORR can be estimated directly, but comparing response rates across cohorts requires caution because cohort membership may be associated with prognosis, tumor type, mutation, prior therapy, and other factors.
No posted formal statistical analyses in the ClinicalTrials.gov record
The ClinicalTrials.gov record states that results are posted, but no formal statistical analyses were posted. Therefore, this page reports the registered endpoints and the posted safety counts reported in the ClinicalTrials.gov record, but it does not assign a hazard ratio, odds ratio, confidence interval, p-value, or treatment-effect estimate to an endpoint for which the ClinicalTrials.gov record contains no such analysis.
6. Planned Analysis
The registry provides three primary objective-response endpoints, and results are posted, but the registry-reported statistical-analysis field contains no formal statistical analyses. Accordingly, the analysis below describes how endpoints of this type are ordinarily summarized and interpreted rather than claiming that a particular statistical test was performed in SUMMIT.
| Primary endpoint | Typical statistical treatment | What can be concluded from the ClinicalTrials.gov record |
|---|---|---|
| Confirmed ORR by independent central review | Estimate the proportion achieving confirmed CR or PR; report an appropriate binomial confidence interval and, when prespecified, compare the observed response rate with a benchmark. | The registry identifies the endpoint, its cohort, review method, response criteria, and time frame. No formal statistical estimate, confidence interval, or p-value is reported in the statistical-analysis field. |
| Confirmed ORR by investigator review | Estimate the proportion achieving confirmed CR or PR using the investigator-assessed binary endpoint; quantify uncertainty with a binomial confidence interval when appropriate. | The registry identifies the endpoint, cervical cancer cohort, review method, response criteria, and time frame. No formal statistical estimate, confidence interval, or p-value is reported in the statistical-analysis field. |
| ORR at first assessment by investigator review | Estimate the proportion with CR or PR at the first scheduled tumor assessment and quantify uncertainty around that proportion. | The registry identifies the first-assessment endpoint and its 8 or 9 week time frame. No formal statistical estimate, confidence interval, or p-value is reported in the statistical-analysis field. |
Why a binomial framework is natural
If an endpoint records whether each participant responded, the observed response count follows the structure of a binomial proportion under the usual independent-observation framework. The key statistical quantities are the number of responders, the number evaluated, and the resulting proportion. Confidence intervals then describe uncertainty in the estimated proportion.
For these endpoints, the central quantity is a response proportion, not a hazard ratio. A response rate describes how many participants meet the specified response criterion; it does not describe how long participants live, how long a response lasts, or whether treatment caused the response.
A confidence interval around an ORR estimate would describe statistical uncertainty around the underlying response proportion under the chosen analysis framework. It would not describe the range of responses an individual participant might experience.
A p-value, if a formal hypothesis test were prespecified and reported, would address compatibility with a null hypothesis under the test's assumptions. It would not measure the size or clinical importance of the response rate.
7. Safety Results
The ClinicalTrials.gov record provides serious adverse events by arm as affected participants over participants at risk. These counts are reported below exactly as reported in the registry.
| Arm | Serious adverse events |
|---|---|
| Neratinib | 144/317 |
| Neratinib + Fulvestrant | 12/45 |
| Neratinib + Paclitaxel | 13/22 |
| Neratinib + Trastuzumab | 45/92 |
| Neratinib + Fulvestrant + Trastuzumab | 28/90 |
| Fulvestrant | 5/7 |
| Fulvestrant + Trastuzumab | 0/7 |
Serious adverse events are a safety outcome and should be analyzed separately from objective response. An affected/at-risk count identifies the observed safety burden in a particular arm, but the ClinicalTrials.gov record does not provide formal statistical comparisons, exposure-adjusted rates, confidence intervals, or p-values for these serious adverse events.
8. Statistical Methods Explained
Why are the primary endpoints binary?
Each registered primary endpoint classifies participants according to whether they achieved complete or partial response under a specified response-assessment framework. That makes the endpoint naturally binary at the participant level: responder versus non-responder.
What does an objective response rate measure?
ORR is the percentage of evaluated participants who achieve complete response or partial response according to the applicable response criteria. It is a tumor-response endpoint. It is not the same as overall survival, progression-free survival, or duration of response.
Why does independent central review matter?
Independent central review creates a response assessment that is separate from the investigator's local assessment. In the SUMMIT registry, the breast cancer primary endpoint specifically uses independent central review, whereas the cervical cancer and all-other-cohort primary endpoints use investigator review. Those are distinct measurement procedures and should not be silently combined.
Why does the 8 or 9 week time frame matter?
The all-other-cohorts primary endpoint is defined at the first scheduled tumor assessment, assessed up to 8 or 9 weeks. A response rate measured at an early fixed assessment is answering a different question from a confirmed response rate assessed from enrollment over a period extending up to 58 months.
Why is non-randomization important for interpretation?
In a randomized trial, treatment assignment helps balance measured and unmeasured prognostic factors in expectation. In a non-randomized study, treatment cohorts can differ systematically. Therefore, a higher observed ORR in one cohort cannot automatically be interpreted as a causal treatment effect attributable to the intervention.
What would a confidence interval around ORR tell us?
A confidence interval would quantify uncertainty around the estimated response proportion under the selected statistical method. A narrow interval would indicate greater statistical precision, while a wider interval would indicate greater uncertainty. The interval would not describe individual-level response variability or establish causality.
9. Review Method and Measurement Issues
Central review
The breast cancer primary endpoint uses independent central review, creating an assessment process distinct from investigator review.
Investigator review
The cervical cancer primary endpoint and the all-other-cohorts first-assessment endpoint use investigator review.
RECIST v1.1
The registry identifies RECIST v1.1 for the specified breast and cervical cancer response endpoints.
RECIST or PERCIST
The all-other-cohorts endpoint permits RECIST v1.1 or other defined response criteria and identifies RECIST, if assessed, or PERCIST.
These measurement differences are not merely administrative details. When the outcome is response status, the statistical analysis begins with how response is defined and observed. A change in assessment time, confirmation requirement, imaging review process, or response criterion can change the resulting binary endpoint even when the underlying participants are the same.
10. Why This Trial Matters Statistically
SUMMIT is a useful statistical teaching case because it illustrates how modern basket studies can generate multiple response questions within one overall protocol. The statistical challenge is not simply calculating a response percentage; it is matching the analysis to the cohort, response definition, assessment method, and time frame.
| Concept | How it appears in SUMMIT |
|---|---|
| Binary endpoint | All three registered primary endpoints are binary objective-response endpoints. |
| Objective response rate | Primary efficacy questions are framed around the percentage achieving CR or PR. |
| Independent central review | Used for the confirmed ORR endpoint in the breast cancer with prior CDK46i cohort. |
| Investigator review | Used for the cervical cancer confirmed ORR endpoint and the all-other-cohorts first-assessment ORR endpoint. |
| RECIST v1.1 | Specified for the breast cancer and cervical cancer response endpoints. |
| PERCIST | Identified as an applicable response framework for the all-other-cohorts endpoint when assessed. |
| Non-randomized design | Treatment assignment was not randomized, limiting straightforward causal interpretation of between-cohort differences. |
| Multiple cohorts | The primary endpoints apply to distinct clinical cohorts rather than one homogeneous randomized population. |
| Time-window definition | Two primary endpoints are assessed up to 58 months, while the all-other-cohorts endpoint is assessed at the first scheduled assessment up to 8 or 9 weeks. |
| Safety denominator | Serious adverse events are reported as affected participants over participants at risk for each registry-reported arm. |
11. Limitations
- Non-randomized allocation: treatment groups are not created by random assignment, so observed differences between cohorts cannot automatically be interpreted as causal treatment effects.
- Multiple primary endpoints: the study has three registered primary objective-response endpoints, each tied to a different cohort and assessment framework.
- Different time frames: two endpoints use a response window assessed up to 58 months, whereas the all-other-cohorts endpoint focuses on the first scheduled tumor assessment up to 8 or 9 weeks.
- Different review methods: independent central review and investigator review can produce different response classifications and should not be treated as identical measurement processes.
- Different response frameworks: the ClinicalTrials.gov record identifies RECIST v1.1 for specified cohorts and RECIST or PERCIST for the all-other-cohorts endpoint.
- No formal statistical-analysis results reported: although results are posted, no formal statistical analyses were posted. Formal effect estimates, confidence intervals, and p-values therefore cannot be attributed to the registry from the ClinicalTrials.gov record.
- Arm-count discrepancy in the registry-reported extract: the ClinicalTrials.gov record states six arms while the serious-adverse-event field contains seven arm labels. The discrepancy is preserved rather than resolved through inference.
- Safety interpretation: serious-adverse-event counts alone do not establish comparative safety without information about exposure, follow-up, event definitions, and a prespecified analysis population.
12. Interpreting Objective Response Carefully
An objective response rate describes the proportion of participants who meet the specified definition of complete or partial response. It is a direct summary of a binary tumor-response endpoint.
ORR does not measure the duration of a response, overall survival, or the probability that treatment caused an individual patient's response. Those questions require different endpoints and analyses.
A response percentage is meaningful only when the population included in its denominator is clearly defined. In SUMMIT, the three primary endpoints differ in cohort, review method, and assessment window. Consequently, an ORR from one primary endpoint should not be casually compared with an ORR from another as if they represented the same statistical population.
The ClinicalTrials.gov record contains no formal statistical-analysis results. That means an observed response rate, even if available elsewhere in the posted outcomes, should not be assigned a statistical significance claim from this record alone. Statistical significance requires a defined null hypothesis, test procedure, analysis population, and error-control framework.
13. Planned Analysis by Endpoint
Confirmed ORR by Independent Central Review
The breast cancer with prior CDK46i cohort uses confirmed ORR by independent central review, with the time frame running from enrollment to the first confirmed CR or PR, whichever came earlier, assessed up to 58 months. The natural statistical summary is the proportion of participants achieving confirmed CR or PR, accompanied by an appropriate binomial confidence interval when inferential precision is required.
The central-review feature is especially important because response status is an assessed outcome rather than a directly measured continuous variable. The statistical estimate inherits the classification process used to determine whether a participant meets RECIST v1.1 response criteria.
Confirmed ORR by Investigator Review
The cervical cancer cohort uses the same broad response concept but with investigator review. The time frame is from enrollment to the first confirmed CR or PR, whichever came earlier, assessed up to 58 months.
The distinction from the breast cancer endpoint is therefore not merely the disease cohort. The review mechanism is also different. Any comparison between the two should account for both clinical-cohort differences and measurement-process differences.
ORR at First Assessment by Investigator Review
The all-other-cohorts endpoint measures CR or PR at the first scheduled tumor assessment, from the first treatment date to the first CR or PR, whichever came earlier, assessed up to 8 or 9 weeks. This endpoint provides an early-response snapshot rather than the same longitudinal confirmation framework used by the first two primary endpoints.
14. Design Topics Not Supported by the ClinicalTrials.gov record
The ClinicalTrials.gov record identifies the allocation, design model, masking status, primary purpose, endpoint types, and posted safety information. They do not provide a formal statistical-analysis record describing a non-inferiority margin, crossover strategy, factorial structure, multiplicity adjustment, interim-analysis boundary, missing-data imputation method, stratification factors, or Bayesian analysis.
| Design topic | What can be stated from the ClinicalTrials.gov record |
|---|---|
| Non-inferiority margin | No non-inferiority margin is reported. |
| Crossover | No crossover analysis is reported. |
| Factorial design | The design is identified as parallel; no factorial structure is reported. |
| Multiplicity | Three primary endpoints are registered, but no formal multiplicity-adjustment procedure is reported. |
| Interim analysis | No interim-analysis method or boundary is reported. |
| Missing-data / imputation | No imputation method is reported. |
| Stratification | No stratification factors are reported. |
| Bayesian methods | No Bayesian method is reported. |
These omissions are important because they prevent a reconstruction of the trial's complete confirmatory statistical plan from the registry extract. A rigorous analysis should distinguish what is documented from what would merely be a conventional methodological assumption.
15. Statistical Concepts in This Trial
Learn more about the methods used in this trial:
16. Related Statistical Calculators
17. Sources
- ClinicalTrials.gov: NCT01953926 — SUMMIT.
- Linked publication: PubMed PMID 41193346.
- Linked publication: PubMed PMID 38211393.
- Linked publication: PubMed PMID 37597578.
- Linked publication: PubMed PMID 36746967.
- Linked publication: PubMed PMID 35181666.
Continue through Clinical Biostats
Use the related tutorials and statistical calculators to explore the methods behind objective response endpoints, binary outcomes, confidence intervals, and clinical-trial analysis.
18. Record Summary
SUMMIT is a phase 2, non-randomized, parallel study of participants with solid tumors harboring somatic HER2 or EGFR exon 18 mutations. Its statistical structure is centered on three binary objective-response endpoints that differ by clinical cohort, response-review method, and assessment time frame. Two endpoints use confirmed response assessed up to 58 months, while the all-other-cohorts endpoint evaluates response at the first scheduled tumor assessment up to 8 or 9 weeks.
The ClinicalTrials.gov record reports 582 enrolled participants, six study arms, three registered primary endpoints, twelve posted outcome measures, and serious-adverse-event counts by the registry-reported arm labels. Because the statistical-analysis field contains no formal analyses, this page deliberately does not create effect estimates, confidence intervals, or p-values that are absent from the ClinicalTrials.gov record.