An Approach-Creating Trial Design Models Jan 2015 Peter Mesenbrink Sangeeta Bhattacharya.

An Approach-Creating Trial Design ModelsJan 2015

Peter Mesenbrink

Sangeeta Bhattacharya

Why is the Trial Design Model Important?

SDTM Trial Design Model is a required part of all CDISC SDTM electronic data submissions to the FDA

The metadata defined for Trial Elements (TE), Trial Arms (TA), Trial Visits (TV), Trial Summary (TS) and Trial Inclusion/Exclusion (TI) is converted into SAS data sets and included as part of the Case Report Tabulations that are part of the eCTD submitted to the FDA

It is used the derivation of other SDTM and ADaM data setsSubject Elements (SE) – Planned and unplanned subject-level trial elements

Subject Visits (SV) – Planned and unplanned subject-level trial visits

Planned and actual treatment groups in ADSL

2 | Overview of Trial Design Model | Business Use Only

Trial Design Model - Purpose

Define the study-level visit structure

Define a subject’s planned path through the study so that study treatment can be packaged and the groups of subjects to be compared can be defined

Provide summary information that is useful in understanding key features of the trial design. This is important in setting up the statistical analysis plan and comparing studies with similar endpoints within a specific disease area.

Overview of TDM

3 | Overview of Trial Design Model | Business Use Only

Subject Elements

Subject Visits

Special Purpose

What “is planned” includes Trial Design domains:

• 5 CDISC domains

What “actually happened” includes 2 Special Purpose domains:

• 2 CDISC subject-level data domains

Trial Design Domain Overview

4 | Overview of Trial Design Model | Business Use Only

Trial Design Domains – What Is Planned

Order of completion

CDISC Domains Description

1 Trial Elements (TE) Basic building blocks for time within the trial. What happens to a subject in each arm (i.e. what series of treatment and non-treatment time periods {trial elements} are planned for a subject assigned to that arm)

2 Trial Arms (TA) Planned sequence of elements, often equivalent to a treatment group.

3 Trial Visits (TV) Planned visits according to the protocol with start and end rules (One record per visit per Arm)

4 Trial Inclusion / Exclusion (TI)

The inclusion/exclusion criteria used to evaluate subjects eligibility for study entry.

5 Trial Summary (TS) Key features of the trial design being implemented / basic sense of what the trial is about. Basic information about the trial such as phase, protocol title, trial objectives, planned and actual # of subjects.

5 | Overview of Trial Design Model | Business Use Only

Special Purpose Domains – What Actually Happened

Domains Description

Subject Elements (SE)

What actually happened to a subject during the study - both what was planned and what was unplanned

Subject Visits (SV)

Actual visits - both those that were planned and those that were unplanned/unscheduled

6 | Overview of Trial Design Model | Business Use Only

Data Flow in Novartis

Study Area

= SDTM+

Much is changing because of CDISC and Novartis Clinical Data Standards (NCDS) – new systems and applications, new / revised processes and roles, etc.

7 | Overview of Trial Design Model | Business Use Only

Challenges with Implementing Trial Design

Every person could have a different interpretation of how to define trial design metadata. Thus as part of project-level planning, teams need to define conventions to be applied across all studies.

Until a fully automated tool is developed there will be some work involved particularly for information that require extraction of text from the protocol and/or is not managed by controlled terminology.

Model will continue to evolve over time as other knowledge is built-up on the availability of searchable metadata within and across clinical trials.

8 | Overview of Trial Design Model | Business Use Only

Trial design model metadata should be a key source in determining the similarity of clinical studies for data pooling

Trial Elements (TE)

ELEMENT and ETCD used in definition of treatment group descriptions in drug packaging and analysis, inconsistency will necessitate mapping for pooled analyses

Start and End rules (TESTRL, TEENRL) – Should be written for easy translation into programmable rules at the subject level in SE. When subjects receive more than one treatment, consistent definitions will simplify how AEs are counted across treatment groups. Also need to ensure that there is no gaps in time when one element ends and the next element starts

Trial Arms (TA)

EPOCH appears in every SDTM data set

There are certain epoch names that are used in Oracle Clinical (OC) that should never appear in the TDM (UNPLANNED and SUMMARY)

ARMCD can be simplified if is difficult to include all treatment elements in 20 characters or less

Areas of focus for TDM consistency across studies (1/2)

Trial Visits (TV)

Planned visit names (VISIT) should be identical between protocol, TDM, and clinical trial database to ensure that all planned and unplanned visits can be merged at the subject level and for easy translation into analysis visits in ADaM

Start rules (TVSTRL, TVENRL) – Need to make clear when assessments taken count towards the planned visit so that unplanned visits can be kept to a minimum

Trial Inclusion/Exclusion (TI)

Does not need to match identically with what is in the protocol (e.g. can remove parenthetical text as needed to get criterion down to 200 characters)

The number of the criterion in the protocol should match the number of the criterion in the TDM (e.g. Inclusion criterion #5 in the protocol, should have IETESTCD = INCL05 in the TDM)

Trial Summary (TS)

Will be updated several times during the course of the study

If TSVAL is not known for a particular TSPARM leave blank and populate TSVALNF accordingly

Areas of focus for TDM consistency across studies (2/2)

Timing of Creation

11 | Overview of Trial Design Model | Business Use Only

Pre Data Build Post Data Build

Pros

Visit structure matches study build

Early attention provided by Study team

Open CDISC run early eliminating late changes to SDTM/ADAM

Pros

No impact on DB build/lock Timelines

Cons

Manage expectations on cross functional Collaboration

Database build timelines impacted

Cons

May Result in Lack of Consistency OpenCDISC run late may force unlocking of

the clinical trial database or revisiting programmed SDTM/ADAM datasets

What have we taken into Account?

Learnings from recent BLAs/NDAs on the content of the SDTM trial design model data sets Supplemental qualifiers not allowed for SDTM trial design model data

sets (i.e. the data elements are fixed and cannot be added to)

Changes and improved understanding of the end-to-end data flow and how the metadata supports it.

To maximize the amount of information in the TDM that can be populated through standard macros and drop down codelists and minimize the amount of manual entry and subsequent re-work

Separate TDM requirements from PK merge requirements

12

How Have we defined a Process

Defined a template (excel spreadsheet with visual basic macros) To be Completed by Statisticians and Programmers

Easy export of different domains to create the necessary SAS data sets

Template Stored in GPS(Unix-Our Statistical Programming Environment)

Simplified versioning and approval process PDF rendition signed by lead statistician and lead statistical programmer

Trial visits to be brought back into Oracle Life Science Hub (LSH) for visit numbering/re-numbering as .csv file with special delimiters

13

What have we defined in terms of Roles & Responsibilities

Shared responsibility of TDM development between trial statistician and trial programmer

Primary accountability is as follows: Trial Visits – Trial programmer in collaboration with Lead Data Manager (LDM) and

Trial Statistician

Trial Elements, Trial Arms, Trial Inclusion/Exclusion, Trial Summary – Trial Statistician (Collaboration with LDM on Trial Inclusion/Exclusion)

Easier export of different domains to create the necessary SAS data sets by Study Programmer

Simplified versioning and approval process Working copy versioned in GPS II /util directory for the clinical study by Study

Statistician

PDF rendition signed by lead statistician and lead statistical programmer

Trial visits to be brought back into LSH for visit numbering/re-numbering as .csv file by Lead Data Manager

14

Other ways that TDM process will hopefully be improved in the near future

Information to be added in disease level/study level analysis plans to define naming conventions for text that is not automated or managed by codelists/controlled terminology Standardization of start and end rules for trial elements and trial visits (e.g. will

the randomization visit be called “RANDOMIZATION” or “BASELINE”, will treatment trial elements always start with the first dose of study treatment?)

Naming conventions for visit names and treatment elements

15

Explanation of the Process

16

17

Instructions for completing the template (1/2)

18

Instructions for completing the template (2/2)

19

Macros tab drives generation of TA and TV (1/2)

20

Macros tab drives generation of TA and TV (2/2)

21

Defining Trial Elements first in TE

22

Instructions for completing the template (1/2)

23

Instructions for completing the template (2/2)

24

Macros tab drives generation of TA and TV (1/2)

25

Macros tab drives generation of TA and TV (2/2)

26

Trial Elements

27

Defining Trial Arms and Epochs

28

TA after providing macro information

29

Defining Trial Visits

30

TV after providing macro information

31

TS and using the controlled terminology

32

The SAS format row tells you the format and length of the value

33

Trial Inclusion/Exclusion simplified

34

TI after extracting information from the protocol

Export TDM

| Presentation Title | Presenter Name | Date | Subject | Business Use Only35

What does the future hold?

| Presentation Title | Presenter Name | Date | Subject | Business Use Only36

Further automated solutions continue to be developed either stand alone or integrated within eProtocol solutions

Challenges remain in having a solution that works in all situations particularly in event-driven and adaptive trial designs but will improve in the future with:Disease-level structured protocols

Therapeutic area standards which will increase the consistency and allow for the development of disease-level trial design model shells