← Clinical Trials
Advanced NSCLC Phase 1 Safety & Tolerability NCT02143466

TATTON: Complete Statistical Analysis of AZD9291 Combinations in Advanced NSCLC

An independent statistical review of the phase 1 TATTON trial evaluating AZD9291 in combination with AZD6094 or selumetinib in patients with advanced non-small-cell lung cancer, with emphasis on its non-randomized dose-development design and safety and dose-limiting-toxicity endpoints.

TATTON  ·  NCT02143466  ·  AstraZeneca  ·  Active, not recruiting
ClinicalTrials.gov record

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

TATTON is a phase 1, non-randomized, parallel-design study of AZD9291 in combination with several novel therapeutics. The registry lists an enrollment of 344 participants, four arms, no masking, and a primary purpose of treatment. The primary registered endpoints focus on adverse events and dose-limiting toxicities as measures of safety and tolerability.

344
Enrollment
Registry enrollment
4
Arms
Parallel design
1
Phase
Phase 1
2014–2020
Study period
Start to primary completion
FeatureTATTON
Study acronymTATTON
NCT identifierNCT02143466
Official titleAZD9291 in Combination With Ascending Doses of Novel Therapeutics
PhasePhase 1
StatusActive, not recruiting
ConditionAdvanced Non Small Cell Lung Cancer
AllocationNon-randomized
Design modelParallel
MaskingNone
Primary purposeTreatment
Enrollment344
Number of arms4
Lead sponsorAstraZeneca
Sponsor typeIndustry
Start date2014-08-05
Primary completion date2020-03-04

2. Clinical Question

The central statistical question in TATTON is not a randomized comparison of efficacy between treatment groups. The registered primary endpoints are safety and tolerability measures for AZD9291 when given in combination with AZD6094 or selumetinib. In Part A, the registry specifically describes the objective of investigating safety and tolerability after progression following prior therapy with an EGFR TKI agent and defining combination dose(s) for further clinical evaluation.

Population

Patients enrolled in a phase 1 study involving advanced non-small-cell lung cancer, with the registry descriptions specifying different populations across study parts.

Interventions

AZD9291 was studied in combination with AZD6094, continuous or intermittent selumetinib, and MEDI4736 across the listed study parts; AZD6094 monotherapy was also included in a Japan-only Part C cohort.

Comparator

The registry identifies the allocation as non-randomized. There is therefore no randomized comparator group for a conventional treatment-effect comparison.

Primary question

How safe and tolerable are the specified combinations, and in Part A what combination dose(s) can be defined for further clinical evaluation?

3. Trial Design

TATTON uses a non-randomized, parallel design without masking. The registry describes four arms and an overall enrollment of 344 participants. The intervention structure is divided into Parts A through D, with different combinations and cohorts appearing within those parts.

01
Enroll 344 participants
02
Study parts Parts A–D
03
Combination dosing Novel therapeutics
04
Safety assessment AE / DLT monitoring
05
Dose definition Further evaluation
Allocation
Non-randomized
Model
Parallel
Masking
None
Purpose
Treatment
Why the non-randomized design matters. Because allocation was non-randomized, the principal statistical task is characterization of safety and tolerability within the studied cohorts rather than estimation of a causal treatment effect from a randomized comparison. Differences between cohorts can reflect differences in the populations enrolled, dose levels, study parts, or treatment combinations as well as differences in observed event rates.

4. Study Parts and Interventions

The registry lists multiple intervention cohorts within the four-arm study structure. These interventions should not be interpreted as four simple randomized treatment groups because the study is explicitly described as non-randomized and the intervention list spans several study parts and cohorts.

PART A

Combination dose investigation

  • AZD9291 in combination with AZD6094
  • AZD9291 in combination with continuous selumetinib in Asian subjects
  • AZD9291 in combination with continuous selumetinib in non-Asian subjects
  • AZD9291 in combination with intermittent selumetinib
  • AZD9291 in combination with MEDI4736
PART B

Combination safety cohorts

  • AZD9291 in combination with AZD6094
  • AZD9291 in combination with selumetinib
  • AZD9291 in combination with MEDI4736
PART C

Japan-only cohorts

  • AZD6094 monotherapy in Japan
  • AZD9291 in combination with AZD6094 in Japan
PART D

AZD6094 combination cohort

  • AZD9291 in combination with AZD6094
  • The registry describes AZD6094 as 300 mg once daily in this part
Interpretation of the arm structure: the intervention list contains more treatment/cohort descriptions than the registry's stated four-arm count. The safest statistical interpretation is therefore to distinguish the study's overall arm designation from its individual intervention cohorts and study parts rather than treating every listed intervention as a separate randomized arm.

5. Primary Endpoints

The registry contains two registered primary endpoint descriptions. Both are safety and tolerability endpoints rather than efficacy endpoints. The first includes dose-limiting toxicities in addition to adverse events and applies to the AZD6094 or selumetinib arms described for Part A. The second focuses on adverse events and describes Parts B, C, and D in addition to the relevant Part A combinations.

Primary endpointTime frameRegistry purpose
Number of participants with Adverse Events and/or Dose Limiting Toxicities as a Measure of Safety and Tolerability of AZD9291 when given in combination with AZD6094 or selumetinib Adverse events will be collected from baseline until 28 days after the last dose in the AZD6094 or selumetinib arms. Part A: investigate safety and tolerability after progression following prior therapy with an EGFR TKI agent, and define combination dose(s) for further clinical evaluation.
Number of participants with Adverse Events as a measure of Safety and Tolerability of AZD9291 when given in combination with AZD6094 or selumetinib Adverse events: baseline until 28 days after the last dose in the AZD6094 or selumetinib arms. Part B, Part C monotherapy and combination cohorts in Japan, and Part D: investigate safety and tolerability in the populations and combinations specified by the registry.

Endpoint structure

The endpoint is fundamentally participant-level safety counting. A participant can be classified according to whether an adverse event and/or dose-limiting toxicity occurred within the registry-defined assessment period. This differs from a time-to-event endpoint such as overall survival: the primary statistical quantity is the number of participants experiencing the defined safety outcome rather than the time until death or progression.

6. Statistical Methodology

Descriptive safety analysis

For a phase 1 safety endpoint defined as the number of participants with adverse events and/or dose-limiting toxicities, the natural primary summary is the number and proportion of participants experiencing the event within each relevant cohort. Because allocation is non-randomized, these summaries are principally descriptive.

Basic safety proportion
p = number of participants with the safety event / number of participants evaluated

The corresponding percentage is obtained by expressing this proportion on a 100-point scale. The denominator should correspond to the relevant analysis population and study cohort rather than automatically using the overall enrollment of 344.

Dose-limiting toxicity assessment

DLTs are especially important in early-phase combination development because they can provide evidence about whether a dose or combination is tolerable enough for further clinical evaluation. The registry's Part A endpoint explicitly combines adverse events and/or DLTs as a measure of safety and tolerability and states that defining combination dose(s) for further clinical evaluation is an objective of that part.

Adverse-event window

The registry specifies that adverse events are collected from baseline until 28 days after the last dose in the AZD6094 or selumetinib arms. This creates a defined observation window for the registered endpoint. Events occurring outside the stated window would not automatically belong to that endpoint simply because they occurred during the broader study.

Why percentages should be paired with counts

Safety percentages can look precise when the underlying cohort is small. Reporting both the number of affected participants and the number evaluated makes the denominator visible and allows readers to judge the amount of information behind the percentage.

No randomized treatment-effect model

The registry identifies the study as non-randomized. A conventional randomized comparison using a treatment-effect estimate such as a hazard ratio would therefore not be the primary framework for interpreting the registered safety endpoints. Cohort-level descriptive summaries are more directly aligned with the stated design and purpose.

7. Statistical Methods Explained

Why is this primarily a descriptive safety analysis?

The primary endpoints count participants with adverse events and, for the Part A endpoint, dose-limiting toxicities. In a phase 1 non-randomized study, these quantities are naturally summarized within the relevant treatment cohorts. The central question is whether the observed safety experience supports continued evaluation of a combination and its dose rather than whether randomization establishes superiority over a control treatment.

Why does the denominator matter for an adverse-event percentage?

A safety percentage is meaningful only when its denominator is clear. The overall enrollment is 344, but the intervention list contains multiple study parts and cohorts. Therefore, a percentage for a particular combination should be understood in relation to the participants actually evaluated for that combination and endpoint.

Why is the 28-day window important?

The registered endpoint specifies adverse-event collection from baseline through 28 days after the last dose in the AZD6094 or selumetinib arms. Defining the observation window prevents an endpoint from silently changing as follow-up continues. It also makes safety summaries more comparable when treatment exposure differs among participants.

What does a dose-limiting toxicity contribute beyond the overall adverse-event count?

A DLT is intended to identify toxicity that is sufficiently important to influence dose development. A broad adverse-event count can include many events of differing severity and clinical significance, whereas a DLT assessment focuses attention on events that are relevant to establishing whether a combination dose can be taken forward.

Why should the four-arm count not be treated as four randomized groups?

The registry explicitly labels allocation as non-randomized. In addition, the intervention list spans Parts A through D and contains several distinct combination and monotherapy cohorts. Consequently, the four-arm designation should not be interpreted as evidence of four randomized treatment groups with exchangeable baseline populations.

Why would a conventional p-value be less central here?

The primary registered question is safety and tolerability in a phase 1, non-randomized setting. A p-value comparing adverse-event rates between non-randomized cohorts would not remove the underlying differences in how participants were allocated. Descriptive event counts, percentages, dose-limiting toxicity assessments, and their clinical context are more directly aligned with the stated study purpose.

8. Planned Analysis

The ClinicalTrials.gov record contains registered primary safety endpoints but no posted statistical analyses. The endpoints therefore define what is to be measured, while the registry does not provide numerical estimates, confidence intervals, or p-values for those endpoints.

Primary endpointTypical analysis for this endpoint typeRegistry result status
Participants with adverse events and/or DLTs Descriptive counts and percentages within the relevant safety population and treatment cohort, with DLTs characterized according to the phase 1 dose-development framework. No statistical analysis posted on ClinicalTrials.gov.
Participants with adverse events Descriptive counts and percentages within the relevant cohort and specified safety observation window. No statistical analysis posted on ClinicalTrials.gov.
What the absence of posted analyses means: the registry identifies safety and tolerability as the primary measurement framework, but it does not provide numerical endpoint results or formal statistical analyses for those endpoints in the record summarized here.

How these endpoints would normally be presented

A complete phase 1 safety analysis would ordinarily present the number of participants evaluated, the number and percentage with at least one adverse event, clinically important event categories, and the number and percentage meeting the DLT definition where applicable. For DLTs, the analysis would also identify the dose level and combination context because dose development is explicitly part of the Part A objective.

Confidence intervals can be added to binomial event proportions when the purpose is to quantify statistical uncertainty. However, the registry information does not provide the event counts or cohort-specific denominators needed to produce such estimates here.

9. Safety Analysis Framework

Safety is the principal statistical domain identified by the registered primary endpoints. The endpoint descriptions specify adverse-event collection through 28 days after the last dose in the AZD6094 or selumetinib arms.

Adverse events

The registry measures the number of participants with adverse events as a safety and tolerability endpoint.

Dose-limiting toxicities

The Part A primary endpoint additionally includes the number of participants with DLTs.

Observation window

Adverse events are collected from baseline until 28 days after the last dose in the AZD6094 or selumetinib arms.

Dose development

Part A specifically aims to define combination dose(s) for further clinical evaluation.

Serious adverse events

The registry information summarized for this analysis does not provide arm-specific counts of serious adverse events. Consequently, no arm-level serious-adverse-event comparison is presented.

10. Analysis Populations and Denominators

The registry gives an overall enrollment of 344 but does not provide cohort-specific analysis populations or participant-level safety counts in the statistical information summarized here. This distinction is important for interpreting future safety tables.

QuantityWhat it tells the reader
Overall enrollmentThe registry states that 344 participants were enrolled in the study.
Number of armsThe registry states that the study has 4 arms.
Intervention cohortsThe registry lists multiple intervention combinations and cohorts across Parts A–D.
Safety denominatorA future adverse-event percentage should use the appropriate participants evaluated for the relevant cohort and endpoint.
DLT denominatorA DLT summary should identify the participants and dose-development cohort to which the DLT assessment applies.

The difference between enrollment and an endpoint-specific denominator is especially important in a multi-part phase 1 study. An overall study enrollment number should not automatically be used to calculate a cohort-specific safety rate.

11. Non-Randomization and Statistical Interpretation

Non-randomization is a defining feature of TATTON. In a randomized trial, allocation creates the foundation for comparing outcome distributions between groups under the randomized treatment assignment. In a non-randomized phase 1 study, that foundation is different: treatment and cohort assignment can be associated with the characteristics of the participants who enter each cohort and with the stage of dose development.

Statistical interpretation

A difference in observed adverse-event rates between two TATTON cohorts would describe a difference in the observed safety experience of those cohorts. It would not, by itself, establish that one combination caused a higher or lower adverse-event probability than another combination because treatment was not randomly assigned.

Why this matters for comparisons

AZD9291 combinations appear in several parts of the study, alongside different cohort descriptions and, in some cases, population distinctions such as Asian versus non-Asian subjects or Japan-only cohorts. Those design features make the clinical context of each safety percentage essential to interpretation.

12. Dose Development and Ascending Doses

The study title describes AZD9291 in combination with ascending doses of novel therapeutics. Part A is explicitly described as investigating safety and tolerability and defining combination dose(s) for further clinical evaluation.

Dose-development logic
Dose level → observed toxicity → tolerability assessment → combination dose selection for further evaluation

The statistical purpose is not simply to count all adverse events. The pattern and severity of toxicity across dose-development cohorts are relevant to identifying a dose that can proceed to additional clinical evaluation.

The registry does not provide enough numerical information in the record summarized here to reconstruct a dose-toxicity curve or estimate a formal dose-response relationship. The appropriate interpretation therefore remains focused on the registered safety and tolerability objectives.

13. Population-Specific Cohorts

The intervention descriptions include population-specific cohorts. In Part A, continuous selumetinib is listed separately for Asian subjects and non-Asian subjects. Part C includes Japan-only monotherapy and combination cohorts.

Study partPopulation distinction stated in the registryIntervention context
Part AAsian subjectsAZD9291 in combination with continuous selumetinib
Part ANon-Asian subjectsAZD9291 in combination with continuous selumetinib
Part CJapan onlyAZD6094 monotherapy
Part CJapan onlyAZD9291 in combination with AZD6094

These distinctions are relevant to statistical reporting because safety summaries should preserve the population and treatment context of the cohort. A pooled percentage can conceal meaningful differences in cohort composition and exposure if the underlying groups are heterogeneous.

14. Time Frame and Censoring Considerations

The registered primary safety endpoint is defined over a finite observation period: from baseline until 28 days after the last dose in the AZD6094 or selumetinib arms. This is fundamentally different from a time-to-event endpoint that follows participants until an event or administrative censoring.

For an adverse-event proportion, the key issue is whether the participant experienced the event during the defined observation window. If follow-up differs because treatment duration differs, the specified window helps establish the endpoint's boundaries. A more detailed exposure-adjusted analysis could use person-time or incidence rates, but those quantities are not part of the registered endpoint descriptions summarized here.

15. Multiplicity and Multiple Cohorts

The registry describes multiple study parts and intervention cohorts, but the registered primary endpoints are safety and tolerability measures rather than a set of randomized confirmatory hypothesis tests. The presence of multiple cohorts therefore should not automatically be interpreted as a multiplicity-adjusted family of efficacy comparisons.

Many cohorts

Multiple combinations and population-specific cohorts create multiple descriptive safety summaries.

Different purpose

Part A specifically includes dose definition for further clinical evaluation, while later parts investigate safety and tolerability in specified populations and combinations.

Where many safety categories or cohorts are examined, isolated statistical significance tests can be difficult to interpret without a prespecified testing framework. For a phase 1 safety-development program, the clinical magnitude, severity, timing, and dose relationship of adverse events can be more informative than a collection of unadjusted p-values.

16. What a Complete Safety Table Should Show

For this type of study, a statistically interpretable safety presentation should make the denominator, event definition, and treatment context explicit.

ElementStatistical purpose
Participants evaluatedDefines the denominator for the safety percentage.
Participants with any adverse eventSummarizes overall participant-level safety burden.
Participants with DLTFocuses on toxicity relevant to dose development where the endpoint applies.
Event countDistinguishes participant-level incidence from the number of individual events.
Study part / cohortPreserves the clinical context of the safety estimate.
Dose levelAllows interpretation of toxicity in the context of dose escalation.
Observation windowDefines which events belong to the registered endpoint.

A crucial distinction is that the "number of participants with adverse events" is not the same as the total number of adverse events. One participant can experience multiple events, while the registered endpoint counts participants experiencing the event.

17. Limitations

18. Why This Trial Matters Statistically

TATTON is a useful teaching case because its statistical structure is different from that of a conventional randomized phase 3 trial. The main analytical challenge is to align the statistical description with a phase 1, non-randomized, multi-cohort safety-development program.

ConceptHow it appears in TATTON
Phase 1 developmentSafety, tolerability, and dose definition are central to the registered primary endpoint framework.
Non-randomizationThe registry explicitly identifies allocation as non-randomized.
Parallel designThe registry identifies the design model as parallel.
Multiple cohortsParts A–D contain several combination and monotherapy cohorts.
Safety endpointParticipants with adverse events are a primary registered outcome.
DLT assessmentPart A includes participants with adverse events and/or DLTs as a primary endpoint.
Defined observation windowAdverse events are collected from baseline until 28 days after the last dose in the AZD6094 or selumetinib arms.
Denominator disciplineSafety percentages must be tied to the appropriate cohort rather than automatically to overall enrollment.
Descriptive statisticsCounts and percentages are the natural primary summaries for participant-level safety endpoints.
GeneralizabilityPopulation-specific cohorts and study-part structure affect how safety observations should be contextualized.

19. Statistical Concepts in This Trial

TATTON illustrates several principles that are important when analyzing early-phase combination studies. The most important is matching the statistical method to the study's actual design and clinical question. A non-randomized safety study should not be interpreted as though it were a randomized efficacy trial.

Participant-level incidence

Counting participants with an adverse event answers a different question from counting the total number of adverse-event occurrences.

Dose-limiting toxicity

DLT assessment connects observed toxicity with the decision about which combination dose(s) should proceed to further clinical evaluation.

Non-randomized comparisons

Observed differences between cohorts do not automatically represent causal treatment effects.

Observation windows

A predefined safety window determines which adverse events contribute to the registered endpoint.

20. Record Timeline

2014-08-05

Study start

The registry lists 2014-08-05 as the study start date.

Phase 1

Combination dose and safety development

The study evaluates AZD9291 in combination with novel therapeutics across Parts A through D, with safety, tolerability, and combination-dose definition forming the registered primary endpoint framework.

2020-03-04

Primary completion date

The registry lists 2020-03-04 as the primary completion date.

Current registry status

Active, not recruiting

The ClinicalTrials.gov record identifies the study status as active, not recruiting.

21. Clinical Interpretation vs Statistical Interpretation

Statistical interpretation

The registered primary endpoints are participant-level safety measures. The study's non-randomized allocation and multi-cohort structure mean that descriptive safety summaries are the central statistical framework.

Clinical interpretation

The registry describes the purpose as investigating safety and tolerability and, in Part A, defining combination dose(s) for further clinical evaluation.

Keeping these two levels separate is important. A statistical safety summary describes what was observed within a cohort; the clinical development question concerns whether that safety experience supports continued investigation of the relevant combination and dose.

22. What the Registry Does and Does Not Report

The ClinicalTrials.gov record supplies the study's design framework and registered primary endpoints, including the adverse-event observation period. It does not provide posted statistical analyses for those primary endpoints in the record summarized here.

Information categoryRegistry information available here
Study designPhase 1, non-randomized, parallel, unmasked, treatment purpose.
Enrollment344 participants.
Arms4 arms.
Intervention structureMultiple combinations and cohorts across Parts A–D.
Primary endpointsAdverse events and, for the Part A endpoint, dose-limiting toxicities.
Safety time frameBaseline until 28 days after the last dose in the AZD6094 or selumetinib arms.
Numerical endpoint resultsNo posted statistical analyses in the record summarized here.
Confidence intervals and p-valuesNo posted statistical analyses providing these quantities.

23. Sources

Continue through the Clinical Biostats trial library

Clinical trial analysis becomes more useful when study design, endpoints, and statistical interpretation are considered together.

24. Record Summary

TATTON is a phase 1, non-randomized, parallel study with 344 participants and four registry-designated arms. Its registered primary statistical focus is safety and tolerability, with adverse events as a primary endpoint and adverse events and/or dose-limiting toxicities included in the Part A endpoint. The registry defines an adverse-event observation period extending from baseline until 28 days after the last dose in the AZD6094 or selumetinib arms. Across Parts A through D, the study includes several AZD9291 combination cohorts, a Japan-only AZD6094 monotherapy cohort, and population-specific selumetinib cohorts.

The key statistical lesson is that the analysis framework should follow the design. Because allocation was non-randomized and the study contains multiple dose-development and population-specific cohorts, safety results are most naturally interpreted through cohort-specific participant counts, percentages, DLT assessment, dose context, and the prespecified observation window. The registry provides the endpoint framework but no posted statistical analyses or numerical primary endpoint results in the record summarized here.

Clinical Biostats methodology: Early-phase clinical trials require a different statistical reading from randomized confirmatory trials. For TATTON, the central task is to understand how safety, tolerability, dose development, cohort structure, and non-randomized allocation determine the appropriate interpretation of the registered endpoints.