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
TRINITI-1 was a completed phase 1/2, non-randomized, single-group study in breast cancer. The registry describes five arms and an enrollment of 104 participants, with ribociclib, everolimus, and exemestane as the study interventions.
| Feature | TRINITI-1 |
|---|---|
| Study acronym | TRINITI-1 |
| Phase | Phase 1/2 |
| Condition | Breast Cancer |
| Population | HR+ HER2- locally advanced/metastatic breast cancer post progression on CDK 4/6 inhibitor |
| Design | Non-randomized, single-group |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 104 |
| Number of arms | 5 |
| Interventions | Ribociclib; Everolimus; Exemestane |
| Status | Completed |
| Lead sponsor | Novartis Pharmaceuticals |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT02732119 |
2. Clinical Question
The registry describes a study of ribociclib with everolimus and exemestane in patients with HR+ HER2- locally advanced or metastatic breast cancer following progression on a CDK 4/6 inhibitor. The primary statistical questions differ between the phase 1 and phase 2 portions: phase 1 focuses on dose-limiting toxicities during the first 28 days, while phase 2 evaluates clinical benefit and best overall response through 24 weeks.
Population
Patients with HR+ HER2- locally advanced/metastatic breast cancer post progression on a CDK 4/6 inhibitor.
Interventions
Ribociclib, everolimus, and exemestane are identified as the study interventions in the registry.
Comparator
The registry describes a non-randomized, single-group design, so there is no randomized comparator group.
Primary questions
In phase 1, how frequently do dose-limiting toxicities occur during cycle 1? In phase 2, does the clinical benefit rate meet the prespecified criterion based on its 95% confidence interval?
3. Trial Design
Five-arm structure
The registry reports five arms. The safety results identify these as Cohort A, Cohort B, Cohort C, Group 1, and Group 2. The record summarized here does not specify a treatment assignment for each of those labels, so the arm names are retained without assigning particular interventions to them.
| Registry arm label | Role in reported safety data |
|---|---|
| Cohort A | Arm-level serious adverse event reporting |
| Cohort B | Arm-level serious adverse event reporting |
| Cohort C | Arm-level serious adverse event reporting |
| Group 1 | Arm-level serious adverse event reporting |
| Group 2 | Arm-level serious adverse event reporting |
4. Primary Endpoints
| Endpoint | Registry definition / assessment | Time frame |
|---|---|---|
| Participants With Dose Limiting Toxicities by Preferred Term in Cycle 1 (28 Days) - in Phase I | A dose-limiting toxicity (DLT) is defined as an adverse event or abnormal laboratory value assessed as having a reasonably possible relationship to the study medication(s) and is unrelated to disease, disease progression, inter-current illness, or concomitant medications that occurs within the first 28 days of treatment (cycle 1) and meets any of the criteria defined in the protocol. National Cancer Institute Common Terminology Criteria for Adverse events (NCI CTCAE) version 4.03 will be used for all grading. | Baseline up to 28 days |
| Clinical Benefit Rate as Per Central Review by Group- Phase II | Clinical Benefit Rate (CBR) is defined as the percentage of participants with a complete response (CR) or partial response (PR) or with stable disease (SD) as per Response Evaluation Criteria in Solid Tumors (RECIST) V1.1 or Non-Complete Response or Non-Progressive Disease (NCRNPD) during first 24 weeks of first dose. CR=disappearance of all non-nodal target lesions, PR=at least a 30% decrease in the sum of diameter of all target lesions, SD=neither sufficient shrinkage to qualify for PR or CR nor an increase in lesions which would qualify for progressive disease (PD), PD=at least a 20% increase in the sum of diameter of all target lesions. The hypothesis was to be rejected if the lower limit of the 95% Confidence Interval for Clinical Benefit Rate (CBR) was greater than 10%. | From baseline up to 24 weeks |
| Best Overall Responses (BOR) Summary Table as Per Central Review by Group - Phase II | Complete response (CR) or partial response (PR), or stable disease (SD) at 24 weeks as per Response Evaluation Criteria in Solid Tumors (RECIST) v1.1 or Non-CR Non-PD (NCRNPD) at Week 24. | Baseline up to 24 weeks and at 24 weeks |
5. Endpoint Architecture
The primary endpoints form a coherent progression from early safety to short-term disease control. Phase 1 asks whether treatment exposure produces protocol-defined DLTs during cycle 1. Phase 2 moves to clinical activity, with central-review clinical benefit measured through the first 24 weeks and a separate best-overall-response summary.
Phase 1: safety signal
DLT is a binary participant-level concept during a fixed 28-day window, but the registry specifies reporting by preferred term. The definition also requires assessment of relationship to study medication(s) and exclusion of several alternative causes.
Phase 2: clinical benefit
CBR combines CR, PR, and qualifying SD or NCRNPD during the first 24 weeks after first dose. It is therefore broader than an objective-response endpoint based only on CR and PR.
Phase 2: best response
BOR summarizes CR, PR, and SD at 24 weeks, with RECIST v1.1 or NCRNPD used for classification.
Central review
Both phase 2 endpoints are explicitly described as being assessed by central review, which reduces dependence on an individual site's response classification.
6. Statistical Methodology
Dose-limiting toxicity analysis
DLT is a fixed-window safety endpoint: participants are evaluated during the first 28 days of treatment, and events qualify only if they satisfy the protocol-defined DLT criteria. Because the endpoint is categorical and the study is non-randomized, the natural statistical summary is the number and proportion of participants experiencing each DLT category, with preferred terms used to describe the observed toxicity pattern.
The denominator must correspond to the prespecified phase 1 analysis population. A participant who does not experience a DLT is not evidence that the treatment has no toxicity; the endpoint describes whether a protocol-defined toxicity occurred within a particular observation window.
Clinical benefit rate
CBR is a proportion endpoint. A participant contributes to the numerator when the central-review assessment satisfies the registry definition of CR, PR, SD, or NCRNPD during the first 24 weeks. The key inferential feature is the prespecified confidence-interval criterion: the hypothesis was to be rejected if the lower limit of the 95% confidence interval for CBR was greater than 10%.
This is a threshold-based design. The statistical question is whether the observed clinical benefit rate is sufficiently above the prespecified 10% benchmark after accounting for uncertainty, rather than whether it differs from a randomized control arm.
Best overall response
BOR is a categorical response summary based on central review. The categories explicitly identified by the registry are CR, PR, and SD at 24 weeks, together with NCRNPD at Week 24. A suitable statistical presentation is a frequency table showing the number and percentage of participants in each response category.
Safety summaries
Serious adverse events are reported at the arm level as affected participants relative to participants at risk. These data are naturally summarized using counts and proportions within each cohort or group. Because there is no randomized comparator, these figures describe observed safety experience within the reported groups rather than a randomized treatment effect.
7. Planned Analysis
The registry marks the three primary endpoints as having results posted. No formal statistical analyses were posted to ClinicalTrials.gov. The statistical framework implied by the endpoint definitions is therefore best understood through the structure of the endpoints themselves.
Phase 1: DLTs
The DLT endpoint would normally be analyzed descriptively within the prespecified phase 1 cohort structure. The principal quantities are the number of participants with a DLT, the corresponding proportion among participants evaluated for the endpoint, and the distribution of DLTs by preferred term. Because the observation window is explicitly limited to baseline through 28 days, events outside that window do not define this primary endpoint unless otherwise specified by the protocol.
Phase 2: Clinical benefit rate
For CBR, the usual analysis begins with the observed proportion achieving CR, PR, SD, or NCRNPD according to the registry definition. A confidence interval for the proportion is then used to address the prespecified threshold. The important feature is the lower confidence limit: the registry states that the hypothesis was to be rejected only when that lower limit was greater than 10%.
Phase 2: Best overall response
For BOR, a categorical summary would normally display each response category together with its frequency and percentage. Because this is a single-group study, the resulting table describes the distribution of observed responses rather than estimating a randomized between-treatment difference.
8. Results Available in the Registry
The ClinicalTrials.gov record marks all three registered primary endpoints as having results posted. The numerical endpoint estimates for DLT, CBR, and BOR are not contained in the record summary represented here, so the available numerical safety information is presented separately below rather than substituting an estimate for an unreported endpoint result.
9. Serious Adverse Events by Arm
The registry reports serious adverse events as the number of affected participants divided by the number at risk in each of the five reported cohorts or groups.
| Arm | Participants with serious adverse events | Participants at risk |
|---|---|---|
| Cohort A | 3 | 8 |
| Cohort B | 3 | 10 |
| Cohort C | 5 | 7 |
| Group 1 | 11 | 46 |
| Group 2 | 5 | 33 |
These counts should be read as arm-level safety summaries, not as evidence of a causal difference between the groups. The denominators are different, and the overall study is non-randomized. In particular, comparing raw serious-adverse-event proportions across these groups would not by itself establish that one group has a higher or lower treatment-related risk.
The reported safety data provide the numerator and denominator needed to describe serious-adverse-event occurrence within each arm. They do not establish a relative risk between arms because participants were not randomly assigned. Differences can reflect the underlying group composition, exposure, sample size, or other design features as well as treatment-related effects.
10. Understanding the 10% Clinical Benefit Threshold
The phase 2 CBR endpoint uses a prespecified threshold rather than a randomized control comparison. The registry states that the hypothesis was to be rejected if the lower limit of the 95% confidence interval for CBR was greater than 10%.
Decision rule
The registry's stated criterion for rejecting the phase 2 hypothesis.
This rule has an important statistical consequence. An observed CBR above 10% is not, by itself, sufficient to satisfy the stated criterion. The estimate must be sufficiently precise that the entire lower confidence bound lies above 10%. A small study can produce a seemingly favorable point estimate but still have a wide confidence interval that crosses the threshold.
The 10% value is a benchmark against which the uncertainty around the single-group CBR is evaluated. It is not a randomized control-group response rate, and the rule does not mean that exactly 10% of participants are expected to benefit.
11. What Clinical Benefit Rate Means
CBR is deliberately broader than a response rate based only on tumor shrinkage. Under the registry definition, clinical benefit includes complete response, partial response, and stable disease, as well as NCRNPD during the first 24 weeks.
| Category | Registry definition | Included in CBR? |
|---|---|---|
| CR | Disappearance of all non-nodal target lesions | Yes |
| PR | At least a 30% decrease in the sum of diameter of all target lesions | Yes |
| SD | Neither sufficient shrinkage to qualify for PR or CR nor an increase in lesions which would qualify for PD | Yes |
| PD | At least a 20% increase in the sum of diameter of all target lesions | No |
| NCRNPD | Non-Complete Response or Non-Progressive Disease | Yes |
The distinction between CBR and objective response is statistically important. A patient does not need to demonstrate tumor shrinkage meeting the PR threshold to contribute to CBR; qualifying stable disease can also count. Consequently, CBR should not be interpreted as the percentage of patients whose tumors shrank.
12. Why Central Review Matters
The phase 2 CBR and BOR endpoints are explicitly defined as being assessed by central review. Central review creates a common response-assessment framework across participants and reduces dependence on individual investigators' assessments.
Consistency
A central reviewer applies the response definitions across the study rather than allowing each participating site to use an independent interpretation as the primary response classification.
RECIST structure
The registry specifies RECIST V1.1 for CR, PR, SD, and PD classification, with NCRNPD also incorporated into the clinical-benefit definition.
Time window
CBR is defined during the first 24 weeks of first dose, while BOR includes the response status at 24 weeks.
Statistical consequence
Response classification is categorical before the CBR proportion is calculated. Misclassification of response can therefore directly change the numerator of the primary phase 2 endpoint.
13. Phase 1 Versus Phase 2 Statistical Questions
Although TRINITI-1 is a single clinical study, its two phases address different statistical problems.
| Feature | Phase 1 | Phase 2 |
|---|---|---|
| Primary focus | Dose-limiting toxicity | Clinical benefit and response |
| Primary observation window | First 28 days | First 24 weeks |
| Endpoint type | Safety event | Categorical efficacy outcome |
| Assessment | Protocol-defined DLT criteria | Central review using RECIST V1.1 or NCRNPD |
| Typical statistical summary | DLT counts and proportions by preferred term | CBR estimate with confidence interval; BOR category distribution |
| Decision framework | Identification of protocol-defined toxicity | Lower 95% CI for CBR compared with 10% threshold |
This distinction prevents a common analytical error: treating all endpoints in a phase 1/2 study as though they were generated by one common hypothesis test. Safety escalation questions and preliminary efficacy questions have different endpoints, time horizons, and inferential objectives.
14. Statistical Methods Explained
Why is DLT measured during the first 28 days?
The registry defines DLT as an event occurring within the first 28 days of treatment, corresponding to cycle 1. A fixed observation window creates a consistent period for evaluating early treatment-related toxicity. It also prevents later adverse events from being silently incorporated into a primary endpoint that was explicitly defined around cycle 1.
Why is CBR a proportion rather than a time-to-event endpoint?
CBR asks whether a participant achieved a qualifying state during a specified 24-week period. The endpoint therefore reduces each participant's response history to a categorical clinical-benefit classification for the defined window. This differs from endpoints such as overall survival, where the time until an event is itself the quantity of interest.
Why is the 95% confidence interval central to the phase 2 decision?
The registry does not define success simply as an observed CBR above 10%. It states that the hypothesis was to be rejected only if the lower limit of the 95% confidence interval was greater than 10%. The confidence interval therefore incorporates sampling uncertainty directly into the decision rule.
What does a 10% threshold mean?
The 10% value is the benchmark specified by the registry for the phase 2 hypothesis. It is not a treatment effect estimate and not a randomized control result. The question is whether the estimated clinical benefit is sufficiently separated from that benchmark given the precision of the estimate.
Why is BOR different from CBR?
CBR is a composite clinical-benefit endpoint evaluated during the first 24 weeks, whereas BOR is a response summary involving CR, PR, and SD at 24 weeks, together with NCRNPD at Week 24. The two endpoints therefore summarize related but different aspects of disease control.
Why do the serious-adverse-event counts not establish a treatment comparison?
The study is non-randomized and single-group. Even when two arms have different observed event proportions, the difference cannot automatically be attributed to treatment because the groups were not created through random allocation. Randomization is what ordinarily provides the foundation for a causal between-group comparison.
Why should confidence intervals accompany a CBR estimate?
A point estimate describes the observed proportion, but it does not describe how precisely that proportion has been estimated. A confidence interval communicates the uncertainty around the estimate and is particularly important here because the prespecified phase 2 decision rule is based on the lower confidence limit.
15. Analysis Population and Non-Randomized Design
TRINITI-1's non-randomized, single-group structure changes the interpretation of nearly every efficacy comparison that might otherwise be made in a randomized trial. There is no randomized treatment contrast from which to estimate a hazard ratio, risk ratio, odds ratio, or mean treatment difference between concurrent arms.
For this reason, an observed clinical benefit rate should not be described as a treatment effect in the same sense as a randomized controlled trial's treatment-versus-control difference. The appropriate interpretation is closer to: the proportion of enrolled participants meeting the prespecified clinical-benefit definition, together with its uncertainty and threshold-based decision rule.
16. Missing Data and Assessment Windows
The endpoint definitions make timing particularly important. DLT is tied to the first 28 days, while CBR and BOR are tied to the first 24 weeks and the Week 24 assessment. Participants with incomplete follow-up can therefore raise questions about endpoint evaluability.
The registry information summarized here does not specify a missing-data or imputation method. For a proportion endpoint, the treatment of participants without an evaluable response assessment can materially affect the estimated CBR. The analysis population and denominator should therefore be defined prospectively rather than selected after observing the response data.
17. Multiplicity and Multiple Primary Endpoints
TRINITI-1 has three registered primary endpoints spanning phase 1 and phase 2: DLT, CBR, and BOR. These endpoints do not represent three interchangeable tests of the same hypothesis. They address different clinical questions at different stages of the study.
| Endpoint | Statistical role | Key issue |
|---|---|---|
| DLT in cycle 1 | Phase 1 safety assessment | Fixed-window toxicity classification |
| CBR through 24 weeks | Phase 2 clinical-benefit assessment | Lower 95% CI compared with 10% threshold |
| BOR through 24 weeks / at 24 weeks | Phase 2 response summary | Categorical distribution of central-review response |
Because the registry's stated CBR rule is explicitly threshold-based, it should not be replaced with an unrelated conventional two-group hypothesis test. Likewise, DLT and BOR should not be treated as though they were additional randomized confirmatory comparisons.
18. Safety Interpretation
The available safety information reports serious adverse events by the five registry arm labels. The counts are small in several groups, making descriptive interpretation more appropriate than treating the arm-level values as precise comparative estimates.
Cohort A
3 of 8 participants were affected by serious adverse events.
Cohort B
3 of 10 participants were affected by serious adverse events.
Cohort C
5 of 7 participants were affected by serious adverse events.
Group 1
11 of 46 participants were affected by serious adverse events.
Group 2
5 of 33 participants were affected by serious adverse events.
The most important statistical feature of these numbers is the denominator. A count such as 3 affected participants has a different evidentiary meaning when the number at risk is 8 than when it is 46. Small denominators also make the corresponding proportions statistically unstable, so apparent differences should not be treated as precise estimates of underlying risk.
19. Why This Trial Matters Statistically
TRINITI-1 is a useful teaching example because it shows how statistical analysis changes when a study moves from early safety assessment to preliminary efficacy assessment without a randomized control group.
| Concept | How it appears in TRINITI-1 |
|---|---|
| Phase 1 safety analysis | DLT assessed during the first 28 days of treatment |
| Categorical endpoint | CBR and BOR classify participants according to predefined response states |
| Confidence interval | CBR uses the lower limit of the 95% CI in its prespecified decision criterion |
| Threshold testing | The phase 2 CBR hypothesis is rejected if the lower 95% CI exceeds 10% |
| Central review | CBR and BOR are assessed by central review |
| RECIST V1.1 | CR, PR, SD, and PD classifications use RECIST V1.1 |
| Single-group design | There is no randomized concurrent comparator |
| Safety denominators | Serious adverse events are reported as affected participants relative to participants at risk |
| Fixed assessment windows | DLT uses 28 days; CBR and BOR use 24 weeks / Week 24 |
20. Important Limitations and Interpretation Issues
- Non-randomized design: without random allocation, observed differences among cohorts or groups cannot automatically be interpreted as causal treatment effects.
- Single-group structure: the phase 2 CBR threshold provides a benchmark-based framework rather than a randomized treatment-versus-control comparison.
- Small arm sizes: the serious-adverse-event denominators range from 7 to 46 participants across the reported cohorts and groups, so several arm-level estimates are based on limited numbers of participants.
- Endpoint-specific windows: DLT is restricted to the first 28 days, whereas CBR and BOR concern the first 24 weeks and Week 24. These endpoints should not be combined into a common time horizon.
- Response classification: CBR incorporates stable disease and NCRNPD, so it should not be interpreted as a tumor-shrinkage rate.
- Central-review dependence: response endpoints depend on the central-review classification specified in the registry.
- Missing-data considerations: the available registry description does not specify an imputation method for missing or unevaluable response assessments.
- Limited inferential detail: the registry marks the primary endpoints as having results posted, but no formal statistical analyses are posted in the statistical-analysis record summarized here.
21. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The study uses a phase-specific framework: early DLT assessment for safety, followed by central-review clinical benefit and response assessment. The phase 2 CBR endpoint has an explicit 10% threshold and confidence-interval decision rule.
Clinical interpretation
The endpoints are designed to describe treatment tolerability and short-term clinical activity in a defined breast-cancer population after CDK 4/6 inhibitor progression. Because the study is non-randomized, the results do not provide the same causal evidence as a randomized treatment comparison.
22. A Practical Statistical Reading of TRINITI-1
A useful way to read the study is to move sequentially through the estimands rather than treating the entire trial as one efficacy test.
Is early treatment exposure associated with protocol-defined DLTs?
The relevant endpoint is participant-level DLT occurrence during baseline through 28 days, with grading according to NCI CTCAE version 4.03 and reporting by preferred term.
How many participants achieve clinical benefit?
CBR includes CR, PR, SD, or NCRNPD during the first 24 weeks of first dose, assessed by central review.
Is the lower confidence limit above the benchmark?
The registry states that the phase 2 hypothesis was to be rejected if the lower limit of the 95% confidence interval for CBR was greater than 10%.
What was the best overall response distribution?
BOR summarizes CR, PR, and SD at 24 weeks, together with NCRNPD at Week 24, using central review.
What serious adverse events were observed?
Serious adverse events are reported by arm as affected participants relative to those at risk, providing a descriptive safety view across the five reported cohorts or groups.
23. The Difference Between a Threshold and a Comparator
One of the most important statistical ideas in TRINITI-1 is the distinction between a historical or prespecified benchmark and a concurrent randomized comparator.
| Question | Threshold-based single-group analysis | Randomized comparison |
|---|---|---|
| Reference point | Prespecified benchmark | Concurrent control group |
| Primary contrast | Observed proportion versus benchmark | Treatment versus control |
| TRINITI-1 phase 2 example | CBR relative to 10% threshold | Not the design used for the phase 2 CBR endpoint |
| Causal interpretation | Limited by absence of randomization | Randomization supports causal comparison under the trial assumptions |
The confidence interval does not solve the absence of randomization. It quantifies statistical uncertainty around the estimated CBR, but it does not account for systematic differences between the enrolled population and the population from which a benchmark may have been derived.
24. Sources
- ClinicalTrials.gov: NCT02732119 — TRINITI-1.
Continue through the Clinical Biostats knowledge graph
Explore the broader collection of clinical trial analyses, statistical tutorials, and quantitative methods used in clinical research.
25. Record Summary
TRINITI-1 illustrates the statistical structure of a non-randomized phase 1/2 clinical study. The phase 1 endpoint uses a fixed 28-day window to identify protocol-defined dose-limiting toxicities, while the phase 2 endpoints use central review to evaluate clinical benefit and best overall response through 24 weeks. The CBR endpoint is particularly instructive because the registry specifies a threshold-based decision rule: the hypothesis was to be rejected when the lower limit of the 95% confidence interval exceeded 10%.
The study's non-randomized, single-group design is central to interpretation. Serious adverse events can be summarized by arm, and clinical benefit can be estimated within the enrolled population, but neither should be treated as a randomized treatment effect. The most informative statistical reading therefore combines the endpoint definition, observation window, confidence-interval framework, response classification, and study design rather than relying on a single numerical summary.