Measuring LVEF regularly can show the rate of decline in left ventricle function for a patient with DMD. Therefore, it was selected as an endpoint for this example to measure whether a course of therapy may slow the progression of heart failure for the patient. The second endpoint chosen, NT-proBNP, is a biomarker that may be used to predict or diagnose heart failure.

This section illustrates example analysis datasets for the following endpoints:

  • Percent change in ejection fraction over time (1 year)
  • Decline in ejection fraction over time (1 year) grouped by a decline greater than -5.0%
  • Percent change in NT-proBNP over time (1 year)

Source Data

The SDTM examples used as the source data are from Section 2.1, Basic Systolic FunctionThe records in the CV dataset for which CVTESTCD = "LVEF_C" and "RVEF_C" were selected for analysis. The records in the LB dataset where LBTESTCD = "BNPPRONT" were used to compute the percent change over time and as a potential covariate. Both SDTM domains are used in one analysis dataset to demonstrate the flexibility of ADaM and to show that the various SDTM domains can be combined into one dataset both as rows and also as covariates.

Example Analysis Datasets

The following tables show the analysis dataset metadata and parameters used for this example.

-

Analysis Datasets

ADSL

The ADSL contains subject characteristics and covariates that are important for analyses. Stratification variables can be created in the ADSL to subset the data. More than one SDTM dataset may be used as input to the ADSL. This is a simplified example ADSL dataset; the ADaMIG should be referenced for additional variables.

For this example:

  • Analysis age (AAGE) was included to provide age with more precision.
  • A flag indicating the use of ACE inhibitors (ACEINHFL) is shown with example derivation from the CM dataset.
  • Body surface area at screening (BSASC) was derived from the VS dataset using height and weight at the screening visit. There are many possible calculations for BSA; for this example, the Du Bois method was used.[11] Note that the screening visit and visit 1 occurred at the same time in this example.

ADSL Variable Metadata

$titleHtml

adsl.xpt

RowSTUDYIDUSUBJIDBRTHDTAAGEAGEAGEUSEXRACETRTSDTRFICDTDTHDTDTHFLTRT01PITTFLACEINHFLHEIGHTSCWEIGHTSCBSASC
1DMD-EFLGEDMD-EF-01-1012010-02-0712.312YEARSMBLACK OR AFRICAN AMERICAN2022-06-162022-06-16

Drug AYY119200.82
2DMD-EFLGEDMD-EF-01-101

2008-05-01

14.114YEARSMASIAN

2022-06-13

2022-06-13

Drug AYY115300.95
3DMD-EFLGEDMD-EF-01-1012003-07-1019.019YEARSMNATIVE HAWAIIAN OR OTHER PACIFIC ISLANDER

2022-07-15

2022-07-15

Drug BYN140451.30
4DMD-EFLGEDMD-EF-01-1011999-01-1523.723YEARSMWHITE

2022-09-06

2022-09-06

2023-11-23

YDrug BYY132421.21
$warningHtml

ADEFNTP

The ADEFNTP includes tests pertinent to analyses. The records from CV (CVTESTCD = "LVEF_C" and "RVEF_C") and the records from LB (LBTESTCD = "BNPPRONT") were selected. Many lab draws may occur throughout the year, but only those linked to a specific visit were included. By subsetting the LB dataset by LBLNKID not missing, laboratory data that aligns with CV data can be filtered out and sorted using USUBJID and VISIT. If LBLNKID is not available or the LB dataset does not contain VISIT, a windowing strategy could be used to select the laboratory draw closest to that visit using date of lab draw (LB.LBDTC.).

TAUGDMDCIS-8 - Getting issue details... STATUS

ADEFNTP Variable Metadata

In this example, variables BASE, CHG, PCHG, and CHGCAT1 were added to facilitate analyses of the efficacy endpoints. the variables SRCDOM, SRCVAR, and SRCSEQ provide traceability to the combined source datasets. The tables below represent parameter value level lists for applicable variables.

Example Parameter [CL.PARAM.ADEFNTP]

Example Parameter Code [CL.PARAMCD.ADEFNTP]

Parameter Value List - ADEFNTP [AVAL]

In this table the ADaM parameters are used rather than the source SDTM data.

Parameter Value List - ADEFNTP [CHGCAT1]

$titleHtml

adefntp.xpt

RowSTUDYIDUSUBJIDAAGEAGEUSEXRACETRT01PTRT01AITTFLHEIGHTWEIGHTBSAACEINHFLPARAMPARAMCDPARAMNAVALAVISITAVISITNVISITADTABLFLBASECHGPCHGCHGCAT1SRCDOMSRCVARSRCSEQ
1DMD-EFLGEDMD-EF-01-10112.3YEARSMBLACK OR AFRICAN AMERICANDrug ADrug AY119200.82YLeft Ventricular Ejection Fraction, Cal (%)LVEF_C167Visit 1 (Baseline)1VISIT 12022-05-16Y67


CVCVTESTCD TAUGDMDCIS-13 - Jira project doesn't exist or you don't have permission to view it. 3
2DMD-EFLGEDMD-EF-01-10112.3YEARSMBLACK OR AFRICAN AMERICANDrug ADrug AY119200.82YRight Ventricular Ejection Fraction, Cal (%)RVEF_C274Visit 1 (Baseline)1VISIT 12022-05-16Y74


CVCVTESTCD7
3DMD-EFLGEDMD-EF-01-10112.3YEARSMBLACK OR AFRICAN AMERICANDrug ADrug AY119200.82YN-Terminal ProB-type Natriuretic Peptide (ng/L)BNPPRONT340Visit 1 (Baseline)1VISIT 12022-05-16Y40


LBLBTESTCD1
4DMD-EFLGEDMD-EF-01-10112.3YEARSMBLACK OR AFRICAN AMERICANDrug ADrug AY132321.08YLeft Ventricular Ejection Fraction, Cal (%)LVEF_C160Visit 6 (1 Year)6VISIT 62023-04-23
67-7-10.447761Decline >=-5.0%CVCVTESTCD11
5DMD-EFLGEDMD-EF-01-10112.3YEARSMBLACK OR AFRICAN AMERICANDrug ADrug AY132321.08YRight Ventricular Ejection Fraction, Cal (%)RVEF_C261Visit 6 (1 Year)6VISIT 62023-04-23
74-13-17.567568Decline >=-5.0%CVCVTESTCD15
6DMD-EFLGEDMD-EF-01-10112.3YEARSMBLACK OR AFRICAN AMERICANDrug ADrug AY132321.08Y

N-Terminal ProB-type Natriuretic Peptide (ng/L)

BNPPRONT3900Visit 6 (1 Year)6VISIT 62023-04-23
408602,150Increase >100 ng/LLBLBTESTCD2

Dataset Debug Message

When the JIRA issue(s) present in the dataset have been resolved, and their resolution has been confirmed by governance, please remove them.

ADEFMRI

The ADEFMRI is an efficacy analysis dataset with only a few records for more complex modeling of changes in ejection fraction, plus the proB-type test results. It is an ADaM Basic Data Structure (BDS) dataset TAUGDMDCIS-15 - Getting issue details... STATUS , with additional variables added from ADEFNTP dataset.

The dataset can can be used for analyses with multiple covariates TAUGDMDCIS-14 - Getting issue details... STATUS by taking the lab tests of interest (PARAMCD = "BNPPRONT") or other analysis datasets, and merging variables of interest with the ADEFNTP dataset. This summary dataset is created from other ADaM datasets and only variables used in the analysis are kept.

ADEFMRI Variable Metadata

$titleHtml

adefmri.xpt

RowSTUDYIDUSUBJIDBRTHDTTRT01PAAGEAGUSEXRACETRT01AITTFLBSAACEINHFLBNPPRONTPARAMPARAMCDPARAMNAVALAVISITAVISITNADTABLFLBASECHGPCHGCHGCAT1
1DMD-EFLGEDMD-EF-01-1012010-02-07Drug A12.3YEARSMBLACK OR AFRICAN AMERICANDrug AY0.82Y40Left Ventricular Ejection Fraction, Cal (%)LVEF_C170Visit 1 (Baseline)12022-05-16Y70


2DMD-EFLGEDMD-EF-01-1012010-02-07Drug A12.3YEARSMBLACK OR AFRICAN AMERICANDrug AY0.82Y40Right Ventricular Ejection Fraction, Cal (%)RVEF_C275Visit 1 (Baseline)12022-05-16Y75


3DMD-EFLGEDMD-EF-01-1012010-02-07Drug A12.3YEARSMBLACK OR AFRICAN AMERICANDrug AY1.08Y900Left Ventricular Ejection Fraction, Cal (%)LVEF_C165Visit 6 (1 Year)62023-04-23
70-5-10.447761Decline >==5.0%
4DMD-EFLGEDMD-EF-01-1012010-02-07Drug A12.3YEARSMBLACK OR AFRICAN AMERICANDrug AY1.08Y900Right Ventricular Ejection Fraction, Cal (%)RVEF_C270Visit 6 (1 Year)62023-04-23
75-5-17.567568Decline >==5.0%
$warningHtml

  • No labels

1 Comment

  1. sources for ADSL variables cannot be ADSL
    ADCVNTP.BRTHDT should be predecessor ADSL.BRTHDT and no derivation is needed if it is from a predecessor
    ADCVNTP.ASEQ is not needed since it is not referenced anywhere in this example
    ADCVNTP.BSA s/b BSAV to be consistent with HEIGHTV/WEIGHTV
    ACEINHFL was added to ADSL - this sohuld be a predecessor - it souhld not be created twice
    ADCVNTP.AVISITN should have source populated
    ADCVNTP.ABLFL - VISIT is VISIT 1 and not 1
    ADCVNTP.BASE - what does "by USUBJID and AVISIT" mean?
    ADCVNTP.CHGCAT1/CHGCAT2 - if these are different PARAMCD which is CHGCAT1 not sufficient?
    ADCVNTP.SRCVAR should be added
    ADCVCMR example 2 - not clear what type of multivariate analysis is supported by this example - more detail is needed
    ADCVCMR example 2 - HEIGHTW or WEIGHTW s/b HEIGHTV or WEIGHTV
    ADCVCMR example 2- the PVLM does not match the data - AVAL is not based on BNPPRONT and CHG is not based on BNPCHG
    Table 1 ARM  - why is a data step needed to subset the data?
    Table 1 ARM  - model has AVISITN=2 and the where clause has AVISITN=6
    Table 2 ARM  - no covariates appear in the model
    Table 2 ARM  - data step subsets ADCVCMR and the proc mixed references ADCVNTP
    Table 3 ARM  - AVISIT GE 6 s/b AVISITN = 6