forked from RConsortium/submissions-pilot1
-
Notifications
You must be signed in to change notification settings - Fork 11
QC Findings
Kangjie Zhang edited this page Apr 12, 2024
·
27 revisions
Please add any discrepencies you found between the original ADaM datasets from the CDISC Pilot and the ones we've programmed in R below.
The R-generated ADSL matches the original ADSL from CDISC pilot data, besides the following mismatches:
- Subject 01-702-1082 has a missing value for BMIBLGR1 in the R-generated ADSL, whilst BMIBLGR1 = "<25" in the original ADSL. This is an issue with the original ADSL, as this subject's BMI at baseline (BMIBL) is missing and therefore the subject shouldn't be assigned a BMI at baseline group.
The R-generated ADAE matches the original ADAE from CDISC pilot data, besides the following mismatches: There is an issue with the original CDISC pilot dataset. ADURN is blank, where AESEQ is (5, 6, 7, 8) for the original CDISC dataset for Subject below:
> adae_orig %>%
filter(USUBJID=='01-716-1418') %>%
select(USUBJID,TRTSDT,ASTDT,AENDT,ADURN,ADURU,AESEQ)
# A tibble: 10 × 7
USUBJID TRTSDT ASTDT AENDT ADURN ADURU AESEQ
<chr> <date> <date> <date> <dbl> <chr> <dbl>
1 01-716-1418 2013-05-05 2013-05-05 2013-05-07 3 DAY 1
2 01-716-1418 2013-05-05 2013-05-05 NA NA NA 2
3 01-716-1418 2013-05-05 2013-05-05 2013-05-07 3 DAY 3
4 01-716-1418 2013-05-05 2013-05-07 NA NA NA 4
5 01-716-1418 2013-05-05 2013-07-01 2013-09-26 NA NA 5
6 01-716-1418 2013-05-05 2013-07-01 2013-10-04 NA NA 6
7 01-716-1418 2013-05-05 2013-07-01 2013-09-26 NA NA 7
8 01-716-1418 2013-05-05 2013-07-01 2013-10-04 NA NA 8
9 01-716-1418 2013-05-05 2013-09-26 2013-11-11 47 DAY 9
10 01-716-1418 2013-05-05 2013-09-26 2013-11-11 47 DAY 10Because it seems the original SDTM.AE.AESTDTC was missing Day, where it seems the original ADAE derivation for ADURN was probably using this date instead of the imputed date. Because day is missing in AESTDTC, ADURN can't derive days.
> ae %>% filter(USUBJID=='01-716-1418') %>% select(USUBJID,AESTDTC,AESEQ) %>% arrange(AESEQ)
# A tibble: 10 × 3
USUBJID AESTDTC AESEQ
<chr> <chr> <dbl>
1 01-716-1418 2013-05-05 1
2 01-716-1418 2013-05-05 2
3 01-716-1418 2013-05-05 3
4 01-716-1418 2013-05-07 4
5 01-716-1418 2013-07 5
6 01-716-1418 2013-07 6
7 01-716-1418 2013-07 7
8 01-716-1418 2013-07 8
9 01-716-1418 2013-09-26 9
10 01-716-1418 2013-09-26 10but the same records, derived in the Pilot 3 dataset do show a calculation since we are using the imputed ASTDT, per the define (ADURN=AENDT-ASTDT+1).
#AE.AESTDTC, converted to a numeric SAS date. Some events with partial dates are imputed in a conservative
#manner. If the day component is missing, a value of '01' is used. If both the month and day are missing
#no imputation is performed as these dates clearly indicate a start prior to the beginning of treatment.
#There are no events with completely missing start dates.
> adae0 %>% filter(USUBJID=='01-716-1418') %>% select(USUBJID,TRTSDT,ASTDT,AESTDTC,AENDT,AEENDY,ADURN,ADURU,AESEQ)
# A tibble: 10 × 9
USUBJID TRTSDT ASTDT AESTDTC AENDT AEENDY ADURN ADURU AESEQ
<chr> <date> <date> <chr> <date> <dbl> <dbl> <chr> <dbl>
1 01-716-1418 2013-05-05 2013-05-05 2013-05-05 2013-05-07 3 3 DAY 1
2 01-716-1418 2013-05-05 2013-05-05 2013-05-05 NA NA NA NA 2
3 01-716-1418 2013-05-05 2013-05-05 2013-05-05 2013-05-07 3 3 DAY 3
4 01-716-1418 2013-05-05 2013-05-07 2013-05-07 NA NA NA NA 4
5 01-716-1418 2013-05-05 2013-07-01 2013-07 2013-10-04 153 96 DAY 6
6 01-716-1418 2013-05-05 2013-07-01 2013-07 2013-10-04 153 96 DAY 8
7 01-716-1418 2013-05-05 2013-07-01 2013-07 2013-09-26 145 88 DAY 5
8 01-716-1418 2013-05-05 2013-07-01 2013-07 2013-09-26 145 88 DAY 7
9 01-716-1418 2013-05-05 2013-09-26 2013-09-26 2013-11-11 191 47 DAY 9
10 01-716-1418 2013-05-05 2013-09-26 2013-09-26 2013-11-11 191 47 DAY 10This latter approach should be the correct approach.
Due to this, we have outlined the expected differences here :
> diffdf(adae, adae_orig, keys = c("STUDYID", "USUBJID", "AESEQ"))
Differences found between the objects!
A summary is given below.
There are columns in BASE and COMPARE with differing attributes !!
All rows are shown in table below- ADURN values will be populated in Pilot 3 (i.e. under
BASE), following the latter derivation approach (i.e. ADURN=AENDT-ASTDT+1) for Subject 01-716-1418 where AESEQ is (5, 6, 7, 8) specified in define.
All rows are shown in table below
===========================================================
VARIABLE STUDYID USUBJID AESEQ BASE COMPARE
-----------------------------------------------------------
ADURN CDISCPILOT01 01-716-1418 5 88 <NA>
ADURN CDISCPILOT01 01-716-1418 6 96 <NA>
ADURN CDISCPILOT01 01-716-1418 7 88 <NA>
ADURN CDISCPILOT01 01-716-1418 8 96 <NA>
------------------------------------------------------------ ADURU should be set to 'DAYS' (i.e. under
BASE) instead of 'DAY' when ADURN is not missing. Updated in Pilot 3 define.
First 10 of 718 rows are shown in table below
===========================================================
VARIABLE STUDYID USUBJID AESEQ BASE COMPARE
-----------------------------------------------------------
ADURU CDISCPILOT01 01-701-1015 3 DAYS DAY
ADURU CDISCPILOT01 01-701-1023 1 DAYS DAY
ADURU CDISCPILOT01 01-701-1023 4 DAYS DAY
ADURU CDISCPILOT01 01-701-1047 1 DAYS DAY
ADURU CDISCPILOT01 01-701-1047 2 DAYS DAY
ADURU CDISCPILOT01 01-701-1097 2 DAYS DAY
ADURU CDISCPILOT01 01-701-1097 3 DAYS DAY
ADURU CDISCPILOT01 01-701-1097 5 DAYS DAY
ADURU CDISCPILOT01 01-701-1097 6 DAYS DAY
ADURU CDISCPILOT01 01-701-1097 7 DAYS DAY
-----------------------------------------------------------The R-generated ADLBC matches the original ADLBC from CDISC pilot data, besides the following mismatches:
Three variables from R-generated ADLBC have class date while the same variables are numeric in the CDISC ADLBC. We opted to keep the date class in our R-generated ADLB.
> diffdf(adlbc, qc_adlbc, keys = c("STUDYID", "USUBJID", "AVISIT", "LBSEQ"))
Differences found between the objects!
A summary is given below.
There are columns in BASE and COMPARE with different classes !!
All rows are shown in table below
==================================
VARIABLE CLASS.BASE CLASS.COMP
----------------------------------
ADT Date numeric
TRTEDT Date numeric
TRTSDT Date numeric
----------------------------------The R-generated ADADAS matches original ADADAS from CDISC pilot data, except for the records where PARAMCD=ACTOT, DTYPE=LOCF. This is an issue from the CDISC ADADAS.
- CDISC SDTM/QS: 818 records for
QSTESTCD=ACTOT - CDISC ADaM/ADADAS: 1040 records for
PARAMCD=ACTOT, 799 (directly from QS, should be 818) + 241 imputed records (DTYPE=LOCF) - ADADAS generated by R: 1040 records for
PARAMCD=ACTOT, 818 (directly from QS) + 222 imputed records (DTYPE=LOCF)
Take a detailed example USUBJID="01-701-1294"
CDISC QS:
> qs %>% filter(QSTESTCD=="ACTOT") %>%
+ select(USUBJID, QSSEQ, VISIT, QSTESTCD, QSTEST,QSSTRESN) %>%
+ filter(USUBJID=="01-701-1294")
# A tibble: 4 × 6
USUBJID QSSEQ VISIT QSTESTCD QSTEST QSSTRESN
<chr> <dbl> <chr> <chr> <chr> <dbl>
1 01-701-1294 5015 BASELINE ACTOT ADAS-COG(11) Subscore 9
2 01-701-1294 5030 WEEK 8 ACTOT ADAS-COG(11) Subscore 14
3 01-701-1294 5045 WEEK 12 ACTOT ADAS-COG(11) Subscore 6
4 01-701-1294 5060 RETRIEVAL ACTOT ADAS-COG(11) Subscore 9CDISC ADADAS:
For the record with QSSEQ=5045 and AVISIT=Week 8, DTYPE is populated as LOCF , but this record is directly from qs dataset, not imputed.
> qc_adadas %>% filter(PARAMCD=="ACTOT") %>%
+ select(USUBJID, QSSEQ, PARAMCD, AVISITN, AVISIT, VISIT, AVAL, DTYPE, ANL01FL, ADT, ADY) %>%
+ arrange(USUBJID, AVISITN) %>% filter(USUBJID=="01-701-1294")
# A tibble: 5 × 11
USUBJID QSSEQ PARAMCD AVISITN AVISIT VISIT AVAL DTYPE ANL01FL ADT ADY
<chr> <dbl> <chr> <dbl> <chr> <chr> <dbl> <chr> <chr> <date> <dbl>
1 01-701-1294 5015 ACTOT 0 Baseline BASELINE 9 "" "Y" 2013-03-24 1
2 01-701-1294 5030 ACTOT 8 Week 8 WEEK 8 14 "" "Y" 2013-05-22 60
3 01-701-1294 5045 ACTOT 8 Week 8 WEEK 12 14 "LOCF" "" 2013-06-14 83
4 01-701-1294 5045 ACTOT 16 Week 16 WEEK 12 14 "LOCF" "Y" 2013-06-14 83
5 01-701-1294 5060 ACTOT 24 Week 24 RETRIEVAL 9 "" "Y" 2013-10-08 199ADADAS generated by R:
DTYPE is not LOCF for the record with QSSEQ=5045 and AVISIT=Week 8, as this record is directly from qs.
> adadas %>% filter(PARAMCD=="ACTOT") %>%
+ select(USUBJID, QSSEQ, PARAMCD, AVISITN, AVISIT, VISIT, AVAL, DTYPE, ANL01FL, ADT, ADY) %>%
+ arrange(USUBJID, AVISITN) %>% filter(USUBJID=="01-701-1294")
# A tibble: 5 × 11
USUBJID QSSEQ PARAMCD AVISITN AVISIT VISIT AVAL DTYPE ANL01FL ADT ADY
<chr> <dbl> <chr> <dbl> <chr> <chr> <dbl> <chr> <chr> <date> <dbl>
1 01-701-1294 5015 ACTOT 0 Baseline BASELINE 9 "" "Y" 2013-03-24 1
2 01-701-1294 5030 ACTOT 8 Week 8 WEEK 8 14 "" "Y" 2013-05-22 60
3 01-701-1294 5045 ACTOT 8 Week 8 WEEK 12 6 "" "" 2013-06-14 83
4 01-701-1294 5030 ACTOT 16 Week 16 WEEK 8 14 "LOCF" "Y" 2013-05-22 60
5 01-701-1294 5060 ACTOT 24 Week 24 RETRIEVAL 9 "" "Y" 2013-10-08 199The same issue occurred for other subjects and resulted in the following discrepancies:
There are rows in BASE that are not in COMPARE !!
First 10 of 33 rows are shown in table below
===========================================
USUBJID PARAMCD AVISIT ADT
-------------------------------------------
01-701-1294 ACTOT Week 16 2013-05-22
01-701-1302 ACTOT Week 16 2013-10-22
01-703-1076 ACTOT Week 16 2013-12-17
01-703-1076 ACTOT Week 24 2013-12-17
01-704-1010 ACTOT Week 24 2014-06-13
01-704-1065 ACTOT Week 16 2013-12-20
01-704-1065 ACTOT Week 24 2013-12-20
01-704-1120 ACTOT Week 16 2014-01-27
01-704-1120 ACTOT Week 24 2014-01-27
01-705-1310 ACTOT Week 16 2013-12-26
-------------------------------------------
There are rows in COMPARE that are not in BASE !!
First 10 of 33 rows are shown in table below
===========================================
USUBJID PARAMCD AVISIT ADT
-------------------------------------------
01-701-1294 ACTOT Week 16 2013-06-14
01-701-1302 ACTOT Week 16 2013-11-05
01-703-1076 ACTOT Week 16 2013-12-24
01-703-1076 ACTOT Week 24 2013-12-24
01-704-1010 ACTOT Week 24 2014-07-09
01-704-1065 ACTOT Week 16 2013-12-24
01-704-1065 ACTOT Week 24 2013-12-24
01-704-1120 ACTOT Week 16 2014-02-03
01-704-1120 ACTOT Week 24 2014-02-03
01-705-1310 ACTOT Week 16 2014-01-23
-------------------------------------------
Not all Values Compared Equal
All rows are shown in table below
=============================
Variable No of Differences
-----------------------------
AVAL 19
CHG 19
PCHG 19
DTYPE 19
----------------------------In the CDISC ADADAS, there are 19 subjects whose records have the incorrect DTYPE=LOCF value instead of the expected missing DTYPE, resulting IN different AVAL/CHG/PCHG values for these subjects.
> diff <- diffdf(adadas, qc_adadas, keys = c("USUBJID", "PARAMCD", "AVISIT", "ADT"))
> count(diff$VarDiff_AVAL, USUBJID)
# A tibble: 19 × 2
USUBJID n
<chr> <int>
1 01-701-1294 1
2 01-701-1302 1
3 01-703-1076 1
4 01-704-1065 1
5 01-704-1120 1
6 01-705-1292 1
7 01-705-1310 1
8 01-708-1347 1
9 01-709-1102 1
10 01-709-1259 1
11 01-710-1045 1
12 01-710-1278 1
13 01-710-1300 1
14 01-710-1315 1
15 01-714-1068 1
16 01-715-1107 1
17 01-716-1373 1
18 01-718-1172 1
19 01-718-1250 1The R-generated ADTTE matches original ADTTE from CDISC pilot data except for minor SAS format discrepancies. Since this adtte was generated in R compared to SAS formats, the columns Type & Length in the define should be sufficient enough to describe the attributes of these variables.
> diffdf(adtte, qc_adtte, keys = c("STUDYID", "USUBJID", "PARAMCD", "SRCDOM", "STARTDT"))
Differences found between the objects!
A summary is given below.
There are columns in BASE and COMPARE with differing attributes !!
First 10 of 20 rows are shown in table below
================================================
VARIABLE ATTR_NAME VALUES.BASE VALUES.COMP
------------------------------------------------
AGE format.sas NULL 3
AGEGR1 format.sas NULL $5
AGEGR1N format.sas NULL 3
EVNTDESC format.sas NULL $25
PARAM format.sas NULL $32
PARAMCD format.sas NULL $4
RACE format.sas NULL $32
RACEN format.sas NULL 3
SAFFL format.sas NULL $1
SEX format.sas NULL $1
------------------------------------------------In pilot3, variable labels were updated per ADaM IG 1.1, which caused some discrepancies with original CDISC pilot data label.
| Dataset | Variable | CDISC pilot data label | Pilot3 label |
|---|---|---|---|
| ADAE | ADURN | Analysis Duration (N) | AE Duration (N) |
| ADURU | Analysis Duration Units | AE Duration Units | |
| AOCCFL | 1st Occurrence of Any AE Flag | 1st Occurrence within Subject Flag | |
| ADADAS | ANL01FL | Analysis Record Flag 01 | Analysis Flag 01 |
| ITTFL | Intent-to-Treat Population Flag | Intent-To-Treat Population Flag | |
| ADTTE | SRCDOM | Source Data | Source Domain |