PhUSE WS4 Subteam Meeting Minutes

March 18-19, 2013

Attendees: See last page for contact information

Co-leads: Susan Kenny, Gail Stoner, with Jingyee Kou, and Stephen Wilson as FDA co-leads.

Meeting Summary

We reviewed the nearly published Submission Data Review Guide (SDRG) and agreed that the Analysis Data Review Guide (ADRG) should follow a similar format. We made the following high level assumptions:

  • All Reviewers will reference both the SDRG and the ADRG.
  • There is a linear flow of file creation process where ADaM domains are created from SDTM.based domains yet we recognize the need to have a section of the ADRG describe any alternate process such as ADaM domains being created from non-SDTM domains.
  • We will develop a ADRG template for a single study rather than for an ISS/ISE in mind.
  • We do not want to repeat important information that is in the clinical study report (CSR) or the statistical analysis plan (SAP) in the ADRG. Instead important information would be called out but links provided to direct reviewer to more information.

Monday Afternoon March 18:

During the first day, we brain stormed about topics that should be discussed in the ADRG. The following categories of topics were discussed:

  • Analysis level information that relates to multiple ADaMdomains such as visit windowing, imputation methods, core variables, differences between SDTM and ADaM variables (e.g. ARM vs. TRTxxP, population flags, baseline values)
  • Topics related to individual ADaM domains
  • Key information from the SAP that relates to content of ADaM domains (notably ADSL)
  • Study level issues such as data cutoff procedure, degree of mis-randomizations or dosing issues, screen failures, etc.)
  • Submission of programs
  • Sponsor variable naming conventions

Timelines and Meeting schedule

The following was presented as our target timelines:

Determine the scope of the ADRG (e.g. analysis datasets, programs, overlap with SDRG) / Complete during CSS
Develop draft ADRG template and completion instructions / 2 months
Vet draft ADRG template and completion instructions / 1 month
Finalize draft ADRG template and completion instructions and develop ADRG examples / 3 months
Release ADRG work package for public comment / 1 month
Finalize ADRG work package and release for use / 1.5 months

We agreed to have meetings every other week. We have a number of EU representatives so we wish to have our meetings at an EU friendly time. This may end up being early for the very few west coast team members.

Tuesday afternoon, March 19

During the second day, we divided into two groups to discuss specific issues related to 1) Study level and analysis level information and 2) Information pertaining to individual ADaM domains. These ideas have not been discussed by the entire group and need to be organized into a coherent outline. However, the rough ideas are as follows.

Study Level and Analysis Level Information

Introduction

  • Same people or different people who use SDTM? Assume reviewers are reading both.

Protocol Design

  • Protocol for SDTM; SAP for ADaM – key information
  • Should we replicate SDRG content? Link to it?
  • Schema plus annotation of ADaM variable interpretation (APERIOD, APHASE, ASPER, etc) and relationship to TRTxxP
  • Graphical presentation may be easier than text
  • What other details would we include from SAP/protocol?
  • Duplication from Trial Summary domain?

Data Section

  • Overview of what data are included (e.g., screen failures dropped from SDTM)
  • Brief introduction to primary/key/important analysis and what types of datasets/analyses support them. Methods
  • Similar to analysis results table – reason for creation of each AD

Random idea list related to analysis level information

  • Windowing
  • Describing intermediate analysis datasets and final dataset
  • Data flow to ADaM – any circularity
  • Data dependency, order of creation
  • Data included – screen failures, cutoffs, etc
  • Missing data – imputation methods in general – link to values of DTYPE
  • Overlap/differences between SDTM and ADaM – pop flags, --BLFL, treatment emergent, ARM vsTRTxxP; ACTARM vsTRTxxA
  • What is found in both places, how are they the same or different
  • Complicated derivations with diagrams
  • Pre-processing before applying derivation rules (e.g., collapsing records)
  • Core variables – vars copied onto other datasets; subject-level, visit/cycle-level
  • Data workarounds and rationale
  • Lookup tables (events [SMQ] or meds of interest); external data sources not in SDTM
  • “Mis-randomizations” – TRTxxPvsTRTxxA (and ARM vs ACTARM)
  • When to include subject IDs?
  • Scoring algorithm? Keep it here vs in SAP vs CSR
  • When are TRTxxP used vsTRTxxA; also how are pop flags used in analysis
  • Answer any obvious questions
  • Why is there limited data for a domain
  • How subjects who switched sites were treated for analysis
  • Might have questions such as “Were there any cases of xxx?” Impact on analysis.
  • General caution about presenting info that overlaps with CSR; might reference CSR instead
  • Overview/index of programs submitted
  • Discussion of programs and whether they are executable
  • Compliance issues

Dataset Level Information

Overview/Intro

  • What is in here?
  • All ADs used for analysis vs subset (describe)
  • Programs creating ADs
  • Analysis results metadata
  • Programs to generate analysis results
  • Data flow, dependencies
  • Flow chart might be helpful. Start with SDTM to final AD. Very useful when you have composite EP or highly summarized datasets.
  • Keep it pretty high level. Not intending to get into detail of which SDTM fed into which ADaM.
  • Should absolutely not go more than one page.
  • Data cutoff, screen failures included, etc
  • General/common info re ADs
  • Split files – what was split, basis for split [This might go below]
  • What ADSL variables were copied to other datasets [Might also go in ADSL section]; key patient-level info
  • Sponsor variable naming conventions

Dataset TOC

Order:

  • Same order as define (current preference of this group)
  • Alpha
  • Some sort of grouping

Columns

  • We’ll provide a default set, but sponsor may add or subtract
  • E.g., might want to do a further breakdown for AEs of special interest, Hy’s Law, etc

Other considerations:

  • How to show split datasets

DSN
Origin of datasets / Might prefer a flowchart
Intermediate vs used for analysis vs both
Exploratory
BDS vs TTE, etc
Safety
Primary EP
Secondary EP
Efficacy
PK/PD
Other

Dataset Detail

DSN description:

This dataset contains …

How you pulled data from multiple sources eg for a composite endpoint

ADSL: These are the variables that we used for stratification or subgroup analysis

Explain any differences with corresponding or related SDTM variables

ADAE: description of variables to support specific events of interest valuable

Caution: Avoid duplication/overlap with SAP content or detailed algorithms in define. Important to ensure consistent definition.

This is a reference data… Briefly describe its content and intended use. Also note where it is (e.g., with programs? With data?)

Detailed definition/algorithm that does not work well in define.

Role of important vars – primary EP, sensitivity, exploratory

Parameter guide to ensure understanding of how to use data

What is relationship between SDTM and AD record count? Were SDTM data excluded and why?

Try guided question approach:

  • Are there any datasets that differ from SDTM?
  • AE: were any std queries used? If yes, provide details
  • ADSL: what grouping and selection variables, any relevant details
  • Any issues with study conduct that impacted AD?

First / Last / Organization / email / Phone
Gail / Stoner / J&J / / 215-793-7565
Susan / Kenny / Amgen / / 919-259-1592
Michael / Frederiksen / Novo Nordisk / / +45 3079938
Albert / Chau / Sarah Cannon Research Institute / / +44 7590 964088
Nancy / Friedland / IBM / / 314 252 5821
Cathy / Bezek / AstellasPharma / / 224-205-8663
Beate / Hientzsch / Accovion / / 0049-6196-7709-406
Lex / Jansen / SAS / / 919-531-9860
Lou / Florio / Purdue / / 203-588-7264
olivier / leconte / Novartis / / 41613240207
Yimei / Wang / BoehringerIngelheim / / 203-798-5623
Victor / Falch / Theorem Clinical Research / / 484-679-3357
Megumi / Fujimoto / Eisai / / 201-364-4817
Chris / Price / Roche / / +44 (0) 1707 365982
Yuguan / Zhao / K&L Consulting Services / / 215-283-6035 x101
Misty / Odle / Lilly / / 317-651-1839
Adam / Huang / Celgene / / 908-723-6731
Sudhir / Singh / EMD Serono / / 781-681-2565
Steven / Light / DataCeutics / / 610-970-2333 x 4723
Steve / Wilson / CDER /
Linda / Collins / PharmaStat / / (510) 501-8738
John / Brega / PharmaStat / / 510-482-2432
Mina / Hohlen / FDA /