Transcript Converting PK Data

Converting
Pharmacokinetic Data
to the PP and PC
Domains
Roger Landau
Principal Consultant, Lincoln Safety Group, Phase Forward
10 Sep 2009
Copyright © 2009 Phase Forward Incorporated. All rights reserved.
Introduction
Support for Pharmacokinetic (PK) data
new in SDTM 3.1.2
PK data combines
•CRF timing variables
•Lab findings
•Derived data
Converting to SDTM can be difficult
Will present my recommendations
Interactive participation encouraged
2
Overview
Typical Pharmacokinetic Data Workflow
Logical Data Model
SDTM Data Model
Recommendations for converting data
•Use of RELREC
3
Typical Pharmacokinetic
Data Workflow
………………………………………………………………………………………………………………………..
4
PK Data
 Consists of two types of data
1) Concentration of drug or metabolite in
bodily fluid/substance
 Analyte – drug or metabolic product whose
concentration is analyzed in biologic matrix
Parent compound or metabolite produced when body
metabolizes compound
 Biological Matrix – fluid or substance in a living
being
Usually plasma or urine
Can also be feces, Cerebrospinal Fluid (CSF)
5
PK Data
2) PK Parameters
• Derived values describing how a compound is
metabolized by the body over time
T1/2: Half-life
AUC: Area Under the Curve
C Max: maximum concentration
T Max: time from drug administration to maximum
concentration
– Many others
–
–
–
–
Phase I Crossover Studies
PK analysis usually in
phase I crossover
studies.
Some studies only
include concentration
No PK parameters
In each period/epoch:
Blood samples taken at
many timepoints
Urine collected over
several intervals
7
Typical PK Data Workflow
Clinical DB
Management
System
CRF
Timepoints
Issues:
• Three different data sources
• Three different departments
• Three different systems
• No data standards
Statistical Analysis
Biostatistics
Merged
Concentration Data
PK Parameters
Concentration
Results
Bioanalytical Lab
8
WinNonLin
Pharmacokineticist
Clinical Study
Report
Logical Data Model
………………………………………………………………………………………………………………………..
9
Logical Data Model
Three entities
•Two SDTM domains
•PC and PP
Relationships
•CRF-Timepoints to
Concentration
CRF Timepoints
PK
PK
PK
Subject No
Period
Timepoint
Date-Time
Sample ID
Sample Condition
Not Done
Vomited?
Concentration
PK
PK
PK
PK
PK
Subject No
Period
Timepoint
Matrix
Analyte
Concentration
Conc Units
Exclude
– One-to-many
•Concentration to PK
Parameters
– Many-to-many
– Requires use of RELREC
in SDTM
PK Parameters
PK
PK
PK
PK
PK
Subject No
Period
Matrix
Analyte
PK Parameter
Value
Units
Exclude
SDTM Data Model
………………………………………………………………………………………………………………………..
11
SDTM PK Data Model
Two domains
•PC – Concentration Data
•PP – PK Parameters
Both are Findings domains
•PC has more timing and qualifier variables
PC and PP related to one another
•But relationship is complex
PC Domain
PC Domain
PC Domain
PP Domain
PP Domain
Recommendations for
Converting PK Data
………………………………………………………………………………………………………………………..
18
Converting PK Data - Recommendations
Source Data
•May be in different formats: XPT, XLS, TXT
•May be one PK file per analyte
•CRF Timepoint and Concentration may be in separate
files
– Merging/joining will probably not go smoothly
• Accession number typos, missing records, etc.
•Usually will need to join and/or concatenate datasets
before converting
Analyte
•Maps to different variables in PP and PC
•Compound/metabolite names may be spelled
inconsistently in different datasets
Converting PK Data - Recommendations
Units
•Can be different for different analytes:
– pg/ml
– ng/ml
•Some PK parameters don’t have units
– No of points
– R-squared
PK Parameter Test Codes
•Develop library of PPTEST and PPTESTCD
•Work with Pharmacokineticists
•Encourage use of standard PPTESTCDs in WinNonLin
Converting PK Data - Recommendations
Status Flags
•Useful for statistical analysis and relationship between
PP and PC
•Obtain flags in WinNonLin output indicating values not
to be used in analysis
•May be difficult to obtain due to WinNonLin
shortcomings
Epoch
•Include Epoch variable in PC and PP
•Represents Period in crossover
•Frequently included in source data
Relating PP to PC
Each PK Parameter in PP calculated from set of
concentrations in PC
PP and PC can be joined to show related values
3.1.2 Implementation Guide
•Analysis datasets may document relationship
OR
•RELREC used to represent how to join PP and PC
RELREC method will be discussed
IG – Section 6.3.10.5
Relating PP Records to PC Records
Reading this section of IG will either
Make you head spin
Put you to sleep
Logical Data Model
 Many-to-Many relationship
• Each PK parameter
calculated from several
concentrations
• Each Concentration used in
the calculation of several PK
parameters
CRF Timepoints
PK
PK
PK
Subject No
Period
Timepoint
Date-Time
Sample ID
Sample Condition
Not Done
Vomited?
Concentration
PK
PK
PK
PK
PK
Subject No
Period
Timepoint
Matrix
Analyte
Concentration
Conc Units
Exclude
 RELREC provides data
needed to perform manyto-many join
PK Parameters
PK
PK
PK
PK
PK
Subject No
Period
Matrix
Analyte
PK Parameter
Value
Units
Exclude
RELREC
Can be used in one of two methods
•Relating Datasets (simpler)
•Relating Records (more complex)
RELREC - Relating Datasets
Can only be used if:
•For all subjects
•All concentrations at all timepoints used for all PK
parameters
PCGPRID, PPGRPID = Matrix + Analyte + Period
This situation usually doesn’t happen
•Some values excluded
Acceptable if agreed that excluded values do not
need to be identified in SDTM datasets
RELREC – Relating Records
 Populate RELREC with records from both PP and
PC
•Use RELID to indicate related records
 IG provides 4 examples (1, 2, 3, 4)
 For each example, 4 possible methods (A, B, C,
D) to populate RELREC
 OMG!
•Less would have been more
RELREC – Relating Records
Recommendations
 Only use Method A
• Use PCGRPID and PPGRPID as IDVARs in RELREC
 Set PPGRPID and PCGRPID to Matrix + Analyte +
Period
• PPGRPID = PPSPEC + PPCAT + EPOCH
• PCGRPID = PCSPEC + PCTEST + EPOCH
RELREC – Relating Records
Recommendations
 Use example 2
method to indicate
concentration values
not used in PK
parameter
calculations.
• Usually due to
insufficient sample or
subject vomited
• Can be obtained from
PCSPCCND and
PCSTAT/PCREASND
RELREC – Relating Records
Recommendations (cont’d)
 Do not indicate concentration values not used to
calculate particular PK parameters
•
•
•
•
•
Examples 3 and 4 in IG
Documents Pharmacokineticist’s analytical methods
Information in WinNonLin
Difficult to obtain and document
Not appropriate in SDTM tabulation datasets
 Has any one used any of these methods?
• Which method was used?
• How did it work out?
Summary
 Converting PK data to PP and PC can be
challenging
• Source data from three different systems
 PC and PP follow Findings model
• Several complexities need to be taken into account
 Recommendations listed for converting source
data to SDTM
 Relating PP to PC using RELREC
• Use Method A and example 2 from IG
Thank You
[email protected]
© 2009 Phase Forward Incorporated. All rights reserved.
32