From SDTM to displays, through ADaM & Analyses Results Metadata, - - PowerPoint PPT Presentation

from sdtm to displays through adam analyses results
SMART_READER_LITE
LIVE PREVIEW

From SDTM to displays, through ADaM & Analyses Results Metadata, - - PowerPoint PPT Presentation

From SDTM to displays, through ADaM & Analyses Results Metadata, a flight on board METADATA Airlines Omar SEFIANI - Stphane BOUGET, Boehringer Ingelheim DH13, PhUSE Barcelona 2016, October, 12 th Outline Background Metadata Driven


slide-1
SLIDE 1

From SDTM to displays, through ADaM & Analyses Results Metadata, a flight on board METADATA Airlines

Omar SEFIANI - Stéphane BOUGET, Boehringer Ingelheim

DH13, PhUSE Barcelona 2016, October, 12th

slide-2
SLIDE 2

Outline

  • Background
  • Metadata Driven Programming
  • Metadata Repository
  • SDTM mapping & derivations
  • ADaM implementation
  • Analysis Results Metadata (ARM)
  • Data exportation / documentation
  • Summary

DH13 - PhUSE - Barcelona 2016 2

slide-3
SLIDE 3

Background

  • Many changes can occur during the conduct of a project:

v standard updates leading to structural changes v new scientific approaches v new regulatory requests

  • Find a solution to:

v Reduce maintenance effort v Increase readability v Ensure consistency between similar implementations => One single program per area dealing with multiple trials simultaneously (centric approach)

DH13 - PhUSE - Barcelona 2016 3

slide-4
SLIDE 4

Metadata Driven Programming (1/2)

DH13 - PhUSE - Barcelona 2016 4

  • Metadata used for:

v Generic SAS programs development v SAS Code generation v Multiple small macros implementation (elementary tasks) v Modular programming using KEYWORDS

  • The concept presented is currently used in a multi-trial respiratory

project v BI legacy data conversion to SDTM v ADaM transformation v Reporting program generation

slide-5
SLIDE 5

Definition of algorithms Identification in the code via meaningful KEYWORDS describing the functionality

  • Flexibility
  • Reduction of

maintenance efforts

  • Centralization and

automatization

Metadata Driven Programming (2/2)

DH13 - PhUSE - Barcelona 2016 5

slide-6
SLIDE 6

DH13 - PhUSE - Barcelona 2016 6

Metadata repository (1/4) MS Excel or any MDR system

ARM

TOC (Table of contents Generator) :

  • Unique display

template

  • Output (TLF)
  • Statistics for

main analyses SDTM / ADAM

  • One sheet per SDTM domain / ADAM

dataset (structure)

  • Differentiation of variable types

(mapped or derived / predecessor or derived)

  • Controlled Terminology embedded

SDTM only

  • Link derived variables to derivation methods

keywords

  • One sheet for all methods (link keyword to SAS

macro)

slide-7
SLIDE 7

Metadata repository (2/4): SDTM DM domain

DH13 - PhUSE - Barcelona 2016 7

slide-8
SLIDE 8

Metadata repository (3/4): ADaM setup dataset

DH13 - PhUSE - Barcelona 2016 8

slide-9
SLIDE 9

Metadata repository (4/4): ARM

DH13 - PhUSE - Barcelona 2016 9

slide-10
SLIDE 10

SDTM mapping and derivations (1/2)

Step 1 : SDTM mapping

  • one-to-one relationship

between one legacy variable and one SDTM variable

  • obtain a bijection between
  • ne raw value and one

SDTM value Step 2 : SDTM derivations

  • after all domains are

created

  • facilitating the interaction

between different domains

  • execution order needed

example : EPOCH variable derived first in SE and then populated in

  • ther domains

SDTM transformation engine :

  • All legacy data converted to SDTM on an ongoing basis
  • Includes additional non-required information by SDTM

(e.g. flags)

DH13 - PhUSE - Barcelona 2016 10

slide-11
SLIDE 11

SDTM mapping and derivations (2/2)

EXCEL METADATA SDTM transformation Definition :

  • Raw data

preparation

  • Domains
  • Mapping
  • Derivations

Legacy raw data SDTM Domains raw data Legacy normalised database structure SDTM Domains final data XPT files D E F I N E SAS SAS Structure check Bijectivity check Data preparation Mapping Derivation Content check SAS Exportation SAS

DH13 - PhUSE - Barcelona 2016 11

slide-12
SLIDE 12

ADaM implementation (1/3)

DH13 - PhUSE - Barcelona 2016 12

Metadata ADaMs

  • Global datasets that contains attributes (study or patient level)
  • Can be used by all subsequent ADaMs
  • Automatically replicated to the main sheet as ‘Assigned’ variables to be extracted

from the define.xml during the creation of the ADS.

Manual entry ADaM dataset ADTARM in ADS plan

slide-13
SLIDE 13

ADaM implementation (2/3)

DH13 - PhUSE - Barcelona 2016 13

Setup metadata datasets

  • Contains algorithms identification
  • Read via a generic macro
  • Used to generate a SAS program creating the corresponding metadata dataset
  • Used as setup dataset during the creation of an ADaM
slide-14
SLIDE 14

ADaM implementation (3/3)

DH13 - PhUSE - Barcelona 2016 14

slide-15
SLIDE 15

Analysis Result Metadata (ARM) (1/2)

ARMs used to:

  • Automatically create the ARM section in the define.xml V2 for ADaMs
  • Dynamically generate a part of SAS macro calls for displays

An output program generator is currently under development to automatically generate the different macro calls for:

  • Data building (ADaM preparation)
  • Analysis & reporting (calculation+output)

DH13 - PhUSE - Barcelona 2016 15

slide-16
SLIDE 16

Analysis Result Metadata (ARM) (2/2)

DH13 - PhUSE - Barcelona 2016 16

SAS MACROS

ARMs

Table Of Content (metadata repository)

  • Output description
  • Statistical part of

ARM

TO C GENERAT OR

  • Data selection
  • Macros

parametrization

slide-17
SLIDE 17

Data exportation / documentation

SDTM

  • Define.xml

generated automatically at each execution

  • .XPTs generated at

the same time

  • Pinnacle 21 checks

performed

  • Bijectivity check
  • Possible data

restriction for the export ADAM

  • Define.xml main

source of metadata for the ADaMs creation

  • .XPTs generated at

the same time

  • Pinnacle 21 checks

performed

  • Analysis Data

Reviewer’s Guide contains partially derived information based on the ADS plan ARM

  • Present in the

define.xml

  • Synchronized with

the displays

  • Standardized

results datasets exported for validation

DH13 - PhUSE - Barcelona 2016 17

slide-18
SLIDE 18

Summary

  • Powerful approach to minimize risks of inconsistencies across different

packages and studies - fits with a project centric approach (multiple trials)

  • Development of programs in a generic manner:

v high level of control needed (user ERRORs and WARNINGs ) v custom code or hardcoding should be avoided v good level of algorithmic and use of complex technical solutions (eg. hashcode, extended attributes, arrays , doSubl, multi-level embedded macros, …)

  • Synergy between standardization and flexibility

v more flexible with data diversity v reduction of implementation time and maintenance facility v improvement in productivity for the creation of similar outputs v facilitates delegation and simplify oversight

DH13 - PhUSE - Barcelona 2016 18

slide-19
SLIDE 19

Thank you

DH13 - PhUSE - Barcelona 2016 19

Omar SEFIANI Stéphane BOUGET

  • mar.sefiani@boehringer-ingelheim.com

stephane.bouget@boehringer-ingelheim.com