How to build ADaM BDS dataset from mock up table
7 likes8,045 views
This document provides instructions for building ADaM basic data structures (BDS) from annotated mock up tables. It discusses how to design mock up tables based on the statistical analysis plan, annotate the tables, create metadata, and then build the ADaM BDS datasets according to the metadata. The process results in analysis-ready ADaM datasets where all numbers in the final report can be calculated with one SAS procedure. An example is provided demonstrating how to annotate a mock up table and extract the necessary variables and parameters to include in the ADaM datasets and metadata.
How to build ADaM BDS dataset from mock up table
- 1. How to Build ADaM Basic Data Structure from Mock Up tables By Kevin Lee Cytel, Inc. 1
- 2. Instruction of Basic Data Structure BDS is the standard domain structure in ADaM. BDS is designed as one or more records per subject per analysis parameter per analysis time point. One of the main purposes of ADaM BDS is analysisready, meaning that all the numbers in the final report should be calculated with one procedure in SAS. The naming convention of BDS is ADxxxxxx. 2
- 3. BDS Structure Subject Identifier Variables Treatment Variables Timing Variables Analysis Parameter Variables Analysis Descriptor Variables Indicator Variables Analysis Enabling Variables Data Point Traceability Variables SDTM Variables 3
- 4. Steps to create ADaM BDS from Mock Up tables Design Mock Up Tables (typically created by Statistician) according to SAP Annotate Mock Up Tables Design Metadata according to Mock Up Tables Create ADaM BDS data sets according to Metadata 4
- 5. Flowchart SAP Mock Up tables Metadata Annotated Mock Up tables SDTM 5 ADaM TFL
- 6. Mock Up table Table 14.4.1 Summary of table of Creatine at baseline (Per Protocol Population) Group 1: Treatment 1 (N=xxx) n Mean Observed Value Creatine Log of Creatine 6 n Group 2: Placebo (N=xxx) Mean Observed Value
- 7. Annotated Mock Up table Table 14.4.1 Summary of table of Creatine at baseline ADLB.AVISIT=‘BASELINE’ (Per Protocol Population) ADLB.PPROTFL=‘Y’ Group 1: Treatment 1 (N=xxx) ADLB.TRTAN = 1 n Creatine where ADLB. PARAMCD=‘CREAT’ Mean Observed Value Count( ADLB. AVAL) MEAN(ADL B.AVAL) Log of Creatine where ADLB. PARAMCD=‘L10CREAT’ 7 Group 2: Placebo (N=xxx) ADLB.TRTAN = 2 n Mean Observed Value Count MEAN(ADLB. (ADLB AVAL) .AVAL)
- 8. New Variables according to annotation Protocol population variable – PPROTFL Baseline – AVISIT, AVISITN Treatment variable – TRTAN, TRTA Parameter Variable – PARAM, PARAMCD Observed Mean Value Variable – AVAL New Parameters according to annotation Creatine – Its paramcd is “CREAT” and its analysis values, AVAL, come from LB.LBSTRESN. Log of Creatine – Its paramcd is “L10CREAT” and its analysis values, AVAL, come from log of LB.LBSTRESN. 8
- 9. Analysis Dataset Metadata Class of Documentation Dataset Dataset Name Dataset Description Dataset Location Dataset Structure Key variables of Dataset ADLB Laboratory analysis data ADLB.xpt one record per subject per parameter per analysis timepoint USUBJID, BDS PARAM, AVISIT 9 ADLB.SAS
- 10. Analysis Variable Metadata including Analysis Parameter Value-Level Metadata Parameter Identifier Variable Name Variable Label Type Format Codelist/ Controlle d Term Source/ Derivation Subject Identifier Variables ** ALL ** STUDYID Study Identifier Char $12. ** ALL ** ADDOMAIN Analysis Domain Char $8. ** ALL ** USUBJID Unique Subject Identifier Char $20. LB.USUBJID ** ALL ** SUBJID Subject Identifier for the Study Char $8. ADSL.SUBJID ** ALL ** SITEID Study Site Identifier Char $10. ADSL.SITEID 10 ADSL.STUDYID ADLB Derived
- 11. Parameter Identifier Variable Name Variable Label Type For mat Codelist/Contr olled Term Source/ Derivation Treatment Variables ** ALL ** TRTA Actual Treatment Group Char $20. ADSL.TRTA ** ALL ** TRTAN Actual Treatment Number Num 8. ADSL.TRTAN Timing Variables ** ALL ** AVISIT Analysis Timepoint Description Char $50. BASELINE VISIT 1 LB.VISIT ** ALL ** AVISITN Analysis Timepoint Number Num 8. 0 = BASELINE 1 = VISIT 1 LB.VISITNUM 11
- 12. Parameter Identifier Variable Name Variable Label Type Form at Codelist/C ontrolled Term Source/ Derivation Analysis Parameter Variables CREAT PARAM Parameter Description Char $100. Creatine( mg/dL) LB.LBTESTCD + unit L10CREAT PARAM Parameter Description Char $100. Log of Creatine( mg/dL) LB.LBTESTCD + unit ** ALL ** Parameter Code Char $8. CREAT LB.LBTESTCD L10CREAT L10CREAT PARAMTYP Parameter Type Char $8. DERIVED CREAT AVAL Analysis Value Num 8. LB.LBSTRESN L10CREAT AVAL Analysis Value Num 8. Log10(LB.LBST RESN) PARAMCD 12
- 13. Parameter Identifier Variable Name Variable Label Type For mat Codelist/Co Source/ ntrolled Derivation Term Indicator Variables ** ALL ** PPROTFL Per Protocol Population Flag Char $1. ADSL.PPROT FL ** ALL ** ABLFL Baseline Flag Char $1. ‘Y’ at ADLB.AVISIT= ‘BASELINE’ Supportive Variables ** ALL ** SRCDOM Source Domain Char $8. LB ** ALL ** SRCVAR Source Variable Char $8. LBSTRESN ** ALL ** SRCSEQ Source Sequence Number Num 8. 13 LB.LBSEQ
- 14. Other Possible Variables Analysis Parameter Variables - BASE, CHG Analysis Descriptor variables – DTYPE Categorical variables - CRIT1 and CRIT1FL Indicator variables - ANL1FL Sample codes using ADaM data set proc sql; **** The Count for and Mean value for Creatine and log of Creatine for protocol population at Baseline; create table line1_1 as select trtan, paramcd, count(aval) as count, mean(aval) as mean from adlb where paramcd in (‘CREAT’, ‘L10CREAT’) and avisit = ‘BASELINE’ and pprotfl = ‘Y’ and aval is not missing group by trtan, paramcd; quit; 14
- 15. Advantages using Annotated Mock Up tables Easy to create the metadata Analysis Dataset metadata Analysis Variable metadata Analysis Parameter Value-level metadata Analysis Results metadata Easy to explain Easy for SAS programmers to follow More visual More accurate number of ADaM data sets 15
- 16. Conclusion ADaM is structured as analysis-ready. In order to build analysis-ready ADaM data sets, SAS programmers need to start from analysis, which are Mock Up tables. From Mock Up tables, SAS programmers figure out what procedures could be used for the analysis. Based on the procedure statements and analyses, SAS programmers annotate the Mock Up tables. The annotations on the Mock Up tables will help SAS programmers to find out what variables are needed in the analyses and eventually what variables and parameters should be created in ADaM data sets. SAS programmers now can create Metadata according to the annotations. According to Metadata, SAS programmers can create ADaM data sets from SDTM data sets. From ADaM data sets, SAS programmers will be able to produce all the results in the Mock Up tables using one proc procedure. 16
- 17. Contact Information Kevin Lee Cytel, Inc. Chesterbrook, PA (610) 994 - 9840 Email:Kevin.lee@cytel.com 17