FIRST. and LAST.

Data Step

Overview

FIRST. and LAST. are automatic SAS variables created during BY-group processing. They are commonly used to derive baseline records, endpoint records, first occurrences, last occurrences, and subject-level summaries. In R, similar functionality is achieved using group_by(), slice(), row_number(), and summarise().

⚠ Behavior Difference

FIRST.variable is 1 for the first observation in a BY group, while LAST.variable is 1 for the last observation in a BY group. In R, equivalent logic is typically implemented using group_by() combined with slice(1) or slice(n()).

Syntax Comparison

SAS

proc sort data=ADLB;
  by USUBJID VISITNUM;
run;

data FIRST_REC;
  set ADLB;
  by USUBJID;
  if first.USUBJID;
run;

library(dplyr)

FIRST_REC <- ADLB %>%
  arrange(USUBJID, VISITNUM) %>%
  group_by(USUBJID) %>%
  slice(1) %>%
  ungroup()

Details

FIRST. and LAST. processing is one of the most important Data Step concepts in clinical programming. It is frequently used to derive baseline values, endpoint values, first adverse events, last assessments, and subject-level summaries.

Example 1 — Keep First Record Per SubjectExample 1

Keep only the first record for each subject.

Input Datasets


USUBJID    VISITNUM    AVAL
CDISC001   1           25
CDISC001   2           30
CDISC001   3           35
CDISC002   1           40
CDISC002   2           45

SAS Code

SAS

proc sort data=ADLB;
  by USUBJID VISITNUM;
run;

data FIRST_REC;
  set ADLB;
  by USUBJID;
  if first.USUBJID;
run;

R Code

library(dplyr)

FIRST_REC <- ADLB %>%
  arrange(USUBJID, VISITNUM) %>%
  group_by(USUBJID) %>%
  slice(1) %>%
  ungroup()

Output Dataset


USUBJID    VISITNUM    AVAL
CDISC001   1           25
CDISC002   1           40

📝 Note: FIRST.USUBJID identifies the first observation within each subject.

Example 2 — Derive Baseline RecordExample 2

Keep the earliest record for each subject and laboratory parameter.

Input Datasets


USUBJID    LBTESTCD    VISITNUM    AVAL
CDISC001   ALT         1           25
CDISC001   ALT         2           30
CDISC001   AST         1           40
CDISC001   AST         2           42
CDISC002   ALT         1           35
CDISC002   ALT         2           50

SAS Code

SAS

proc sort data=ADLB;
  by USUBJID LBTESTCD VISITNUM;
run;

data BASELINE;
  set ADLB;
  by USUBJID LBTESTCD;
  if first.LBTESTCD;
run;

R Code

library(dplyr)

BASELINE <- ADLB %>%
  arrange(USUBJID, LBTESTCD, VISITNUM) %>%
  group_by(USUBJID, LBTESTCD) %>%
  slice(1) %>%
  ungroup()

Output Dataset


USUBJID    LBTESTCD    VISITNUM    AVAL
CDISC001   ALT         1           25
CDISC001   AST         1           40
CDISC002   ALT         1           35

📝 Note: A common clinical programming use case is deriving baseline records using FIRST. processing.

Example 3 — Keep Last Record Per SubjectExample 3

Keep only the last record for each subject.

Input Datasets


USUBJID    VISITNUM    AVAL
CDISC001   1           25
CDISC001   2           30
CDISC001   3           35
CDISC002   1           40
CDISC002   2           45

SAS Code

SAS

proc sort data=ADLB;
  by USUBJID VISITNUM;
run;

data LAST_REC;
  set ADLB;
  by USUBJID;
  if last.USUBJID;
run;

R Code

library(dplyr)

LAST_REC <- ADLB %>%
  arrange(USUBJID, VISITNUM) %>%
  group_by(USUBJID) %>%
  slice(n()) %>%
  ungroup()

Output Dataset


USUBJID    VISITNUM    AVAL
CDISC001   3           35
CDISC002   2           45

📝 Note: LAST.USUBJID identifies the final observation within each subject.

Example 4 — Derive Latest AssessmentExample 4

Keep the most recent record for each subject and laboratory parameter.

Input Datasets


USUBJID    LBTESTCD    VISITNUM    AVAL
CDISC001   ALT         1           25
CDISC001   ALT         2           30
CDISC001   ALT         3           35
CDISC001   AST         1           40
CDISC001   AST         2           42
CDISC002   ALT         1           35
CDISC002   ALT         2           50

SAS Code

SAS

proc sort data=ADLB;
  by USUBJID LBTESTCD VISITNUM;
run;

data LAST_ASSESS;
  set ADLB;
  by USUBJID LBTESTCD;
  if last.LBTESTCD;
run;

R Code

library(dplyr)

LAST_ASSESS <- ADLB %>%
  arrange(USUBJID, LBTESTCD, VISITNUM) %>%
  group_by(USUBJID, LBTESTCD) %>%
  slice(n()) %>%
  ungroup()

Output Dataset


USUBJID    LBTESTCD    VISITNUM    AVAL
CDISC001   ALT         3           35
CDISC001   AST         2           42
CDISC002   ALT         2           50

📝 Note: LAST. processing is commonly used to derive endpoint and latest assessment records.

Conclusion

Key Takeaway: • FIRST.variable identifies the first observation in a BY group • LAST.variable identifies the last observation in a BY group • slice(1) and slice_head(n=x) in R corresponds to FIRST. • slice(n()) and slice_tail(n=x) in R corresponds to LAST. • Common clinical uses include baseline and endpoint derivations