Eleni Frangou1, Jane Holmes1, Sharon Love1, Lang’o Odondi1,2,
Claire Hamill3, Naomi McGregor3, Maria Hawkins2 1 Centre for Statistics in Medicine (CSM), NDORMS, The
University of Oxford
2 Department of Oncology, The University of Oxford 3 Oncology Clinical Trials Office (OCTO), The University of
Oxford
25th August 2015
36th ISCB Annual Conference, Utrecht, The Netherlands
Practical considerations in designing a
phase I Time to Event Continual
Reassessment Method (TiTE CRM) trial
in a grant funded CTU
Overview
•
Description of the clinical trial
•
Design
– Time to Event Continual Reassessment Method (TiTE CRM)
•
Set up
•
Future Steps
2
Communication with the Chief
Investigator
Familiarisation with the TiTE
CRM
Statistical
Description of the clinical trial
•
Phase I, single arm, open-label, multicentre, 2 stage trial in
oesophageal cancer
– Stage A – Palliative setting: Radiation and escalating doses of VX970
– Stage B – Radical setting: Definitive chemoradiotherapy using radiotherapy,
in combination with Cisplatin, Capecitabine and escalating doses of VX970
•
Primary objective in each stage is to determine the safety, toxicity
profile and Maximum Tolerated Dose (MTD)
•
The MTD from Stage A will be used to inform the starting dose of
Design of the clinical trial
•
Continual Reassessment Method (CRM)
– Model-based method for finding the MTD which causes a
dose limiting toxicity (DLT) for a specified target toxicity level
– Larger number of patients are treated on or at adjacent
doses of the MTD than the conventional algorithm-based methods
– A priori a relationship between the dose levels and the
toxicity levels is defined namely the Dose –Toxicity Curve (DTC) which is re-evaluated every time a new patient is observed
– Examples of the DTC include the power curve and the
logistic model
– Monotonicity assumption states that toxicity increases with
increasing dose and that efficacy also increases with increasing dose
Design of the clinical trial
•
Time to Event Continual Reassessment Method
(TiTE-CRM)
– Modified version of the CRM
– It accounts for the time to event of possible late onset
toxicities by considering a weighted DTC
– Uses all accumulated information to decide which dose to
assign the next patient
– Results in shorter study duration as it is not necessary for
a patient to be observed for the full observation period before recruiting the next patient
Communication with the Chief Investigator
(CI)
•
Various meetings and email exchanges were required to establish:
– Timelines
– Meaningful prior probabilities (skeleton)
– An adequate number of nested treatment schedules – Target toxicity levels for both stages
– Stopping rules due to success and safety – Escalation rules
– Utilising the data from Stage A for Stage B
•
Simulation results from different scenarios enabled us to make
decisions between different options
Familiarisation with the TiTE CRM
•
This task consisted of:
– Literature review – Self study
– Attending seminars and courses
– Exchanging views and knowledge amongst statisticians
– Marrying up the expectations of the Chief Investigator and the TiTE CRM design
•
External advice was sought through the Methodology Advisory
Statistical Programming and Simulations
•
Statistical package used: R
•
Programmes were developed independently by
two statisticians
– Executable, simulator and processing programmes
developed
•
Existing R functions could not accommodate some
trial specific characteristics
– Pause in recruitment
– Time-to-toxicity distribution – Dose allocation
– Stopping rules
•
Additional time was allowed for debugging and
validating
Statistical Programming and Simulations
•
Extensive simulations were performed to assess the behaviour and
robustness of the model
•
Data were generated under a number of scenarios reflecting a
variety of realistic and extreme cases
•
Each trial was simulated using the actual study’s parameters and
characteristics
•
The simulation results were used to inform different aspects:
– Sample Size – Trial Duration
– Power to detect the MTD – Stopping due to safety
Statistical Programming and Simulations
Simulation Example
Statistical Programming and Simulations
Trial Management
•
Protocol development
– Time delay between idea and protocol development – Transition between Stage A and B
– Dose levels and schedule of events discussions were driven by an ongoing First In
Human Phase I study
– It has received external review and input at a European Cancer Organisation
(ECCO) workshop
•
High staff requirements
– Data input by sites needs to be swift to allow the statisticians to advise on dose – Lack of knowledge in advance of when teleconferences will be required to advise
on dose escalation
– Risk associated with the combined trial interventions is high and the trial will
require intense central monitoring
Costing and Funding
•
The total cost of this trial was estimated to be
higher than
conventional trials due to
– The complexity of the design
– The extra statistical input to design and during the trial – Our unit’s lack of experience in these methods
– The estimated trial duration being 4 years once recruitment starts
•
The trial was costed on the basis of both the average duration and
number of patients and the maximum duration and number of
patients
•
Funding applications were submitted to multiple sources
– New Agents Committee (NAC), Cancer Research UK – Pharmaceutical Company providing the ATR inhibitor – Department of Oncology, The University of Oxford, UK
Set up of the clinical trial
Communication with the CI; 10.9
Trial Design Characteristics; 8.2
Familiarisation with TiTE CRM; 12.5 Statistical Programming; 13.6 Simulations, Results Processing, Debugging; 32.4
Study Dissemination; 7.1 Trial Management; 10.4
Funding and Finance; 4.9
Time (%)
14 Percentages show the statisticians’ time spent on each aspect of the set up process
Future Steps
•
Protocol
•
Statistical Analysis Plan (SAP)
•
Case Report Forms (CRFs) Review
– Data Management Plan – Critical points review
•
Charters
– Data Monitoring Committee – Trial Steering Committee
•
Statistical programmes
– Statisticians should be at the ready to analyse the available data once
there is a new recruit
– Two independent programmes will be used in addition to the titecrm function in
the dfcrm package by Cheung (2013)
•
Trial Master File
Conclusions
•
Significantly more time, man power and resources
than an
algorithm-based trial
were invested during set up
•
We feel we are up to date with the literature and have a good
knowledge of existing model-based methodologies
•
Generic CRM and TiTE CRM analysis and simulation programmes
have been set up and are ready to use in future trials
•
Having explored the behaviour and performance of this
methodology puts us in an advantageous position in designing
more CRM and TiTE CRM trials
Acknowledgments
•
This work is supported by Cancer Research UK (CRUK) trial number
CRUKD/15/011
•
Vertex Pharmaceuticals
•
University of Oxford
•
CRUK/MRC Oxford Institute for Radiation Oncology
•
Oxford Clinical Trials Research Unit (OCTRU)
– Oncology Clinical Trials Office (OCTO) – Centre for Statistics in Medicine (CSM)
•
Oxford University Hospitals NHS Trust (Churchill Hospital)
•
Leeds Teaching Hospitals NHS Trust
•
University of Leeds
Acknowledgments
•
Chief Investigator:
Maria Hawkins
•
Trial Management:
Claire Hamill and Naomi McGregor
•
Clinical Trials Unit Statisticians:
Jane Holmes, Lang’o Odondi and Sharon
Love
•
Oncology Trials Director:
Tim Maughan