1.4.15. Recoding Variables in SAS vs R equivalent

1.4. Migrating from SAS to R: A Skill Conversion Guide

1.4.15. Recoding Variables in SAS vs R

1. Introduction

Recoding variables is a common data management task in clinical trials, especially for SDTM domains such as AE (Adverse Events), DM (Demographics), and LB (Laboratory). Recoding allows you to:

Collapse categories (e.g., recode AESEV from "MILD"/"MODERATE"/"SEVERE" to "NON-SERIOUS"/"SERIOUS")
Reverse code or group lab results (e.g., recode LBORRES to "LOW"/"NORMAL"/"HIGH")
Assign new values for analysis or reporting

It is best practice to create new variables for recoded values to preserve the original data.

2. Basic Recoding: SAS vs R

Task	SAS	R (base/car)
Recode with IF/THEN	`IF AESEV="SEVERE" THEN AESER="Y";`	`ifelse(AESEV == "SEVERE", "Y", "N")`
Recode with array/loop	`ARRAY aesev{*} AESEV1-AESEV4; ...`	`lapply()` or `mutate(across())`
Recode with value labels	`PROC FORMAT`	`factor()` or `recode()` (car package)

3. Input Example (SDTM AE Domain)

Suppose you have the following AE data:

USUBJID	AEDECOD	AESEV	AESER
01-001	HEADACHE	MILD	N
01-002	NAUSEA	MODERATE	N
01-003	DIZZINESS	SEVERE	Y
01-004	FATIGUE	SEVERE	Y
01-005	NAUSEA	MILD	N

4. Recoding a Single Variable

SAS Example:

data ae;
  set ae;
  if AESEV = "SEVERE" then AESER = "Y";
  else AESER = "N";
run;

Recodes AESER: sets to "Y" if AESEV is "SEVERE", otherwise "N".

R Example (base):

# Dummy AE data
ae <- data.frame(
  USUBJID = c("01-001", "01-002", "01-003", "01-004", "01-005"),
  AEDECOD = c("HEADACHE", "NAUSEA", "DIZZINESS", "FATIGUE", "NAUSEA"),
  AESEV = c("MILD", "MODERATE", "SEVERE", "SEVERE", "MILD"),
  AESER = c("N", "N", "Y", "Y", "N")
)

ae$AESER <- ifelse(ae$AESEV == "SEVERE", "Y", "N")

Uses ifelse() for recoding.

R Example (car package):

library(car)
# Dummy AE data already defined above
ae$AESEV_REC <- recode(ae$AESEV, "'MILD'='NON-SERIOUS'; 'MODERATE'='NON-SERIOUS'; 'SEVERE'='SERIOUS'")

Collapses AESEV into "NON-SERIOUS" and "SERIOUS".

Expected Output:

AESEV	AESER	AESEV_REC
MILD	N	NON-SERIOUS
MODERATE	N	NON-SERIOUS
SEVERE	Y	SERIOUS
SEVERE	Y	SERIOUS
MILD	N	NON-SERIOUS

5. Recoding Multiple Variables (e.g., LBORRES in SDTM LB Domain)

Suppose you want to recode multiple lab result columns (e.g., LBORRES1, LBORRES2) to "LOW", "NORMAL", "HIGH" based on reference ranges.

SAS Example:

data lb;
  set lb;
  array lborres{2} LBORRES1 LBORRES2;
  array lbnorm{2} LBNORM1 LBNORM2;
  do i = 1 to 2;
    if lborres{i} < 70 then lbnorm{i} = "LOW";
    else if lborres{i} > 110 then lbnorm{i} = "HIGH";
    else lbnorm{i} = "NORMAL";
  end;
run;

R Example (vectorized):

# Dummy LB data
lb <- data.frame(
  LBORRES1 = c(65, 90, 115),
  LBORRES2 = c(120, 80, 60)
)

lb$LBNORM1 <- ifelse(lb$LBORRES1 < 70, "LOW",
                ifelse(lb$LBORRES1 > 110, "HIGH", "NORMAL"))
lb$LBNORM2 <- ifelse(lb$LBORRES2 < 70, "LOW",
                ifelse(lb$LBORRES2 > 110, "HIGH", "NORMAL"))

Or, for many columns:

lab_cols <- c("LBORRES1", "LBORRES2")
lb[paste0("LBNORM", 1:2)] <- lapply(lb[lab_cols], function(x)
  ifelse(x < 70, "LOW", ifelse(x > 110, "HIGH", "NORMAL")))

Expected Output:

LBORRES1	LBORRES2	LBNORM1	LBNORM2
65	120	LOW	HIGH
90	80	NORMAL	NORMAL
115	60	HIGH	LOW

6. Recoding Continuous to Categorical (e.g., Age Groups in DM Domain)

R Example:

# Dummy DM data
dm <- data.frame(
  AGE = c(12, 34, 70)
)

dm$AGEGRP <- cut(dm$AGE, breaks = c(0, 18, 65, Inf),
                 labels = c("Child", "Adult", "Senior"), right = FALSE)

Categorizes AGE into "Child", "Adult", "Senior".

Expected Output:

AGE	AGEGRP
12	Child
34	Adult
70	Senior

7. Reverse Coding (e.g., Questionnaire Scores)

Suppose QSVAL is a 1–5 scale, and you want to reverse it.

R Example:

# Dummy QS data
qs <- data.frame(
  QSVAL = 1:5
)

qs$QSVAL_REV <- 6 - qs$QSVAL

1→5, 2→4, 3→3, 4→2, 5→1

Expected Output:

QSVAL	QSVAL_REV
1	5
2	4
3	3
4	2
5	1

8. Beyond Basics: Advanced Recoding

Multiple recodes at once: Use dplyr::mutate(across(...)) for many columns.
Custom functions: Write your own recoding logic for complex SDTM domains.
Factor recoding: Use forcats::fct_recode() for factor levels (e.g., DM$SEX).

R Example: dplyr/across

library(dplyr)
# Dummy AE data already defined above
ae <- ae %>%
  mutate(across(starts_with("AESEV"), ~recode(., "'MILD'='NON-SERIOUS';'MODERATE'='NON-SERIOUS';'SEVERE'='SERIOUS'")))

R Example: forcats

library(forcats)
# Dummy DM data for SEX
dm <- data.frame(
  SEX = c("F", "M", "F", "M")
)
dm$SEX <- fct_recode(dm$SEX, Female = "F", Male = "M")

Expected Output:

SEX	recoded_SEX
F	Female
M	Male

9. Input and Output Table: Recoding Example (AE Domain)

Input Table:

USUBJID	AEDECOD	AESEV
01-001	HEADACHE	MILD
01-002	NAUSEA	SEVERE

R Recoding:

# Dummy AE data for recoding example
ae <- data.frame(
  USUBJID = c("01-001", "01-002"),
  AEDECOD = c("HEADACHE", "NAUSEA"),
  AESEV = c("MILD", "SEVERE")
)

ae$AESER <- ifelse(ae$AESEV == "SEVERE", "Y", "N")

Output Table:

USUBJID	AEDECOD	AESEV	AESER
01-001	HEADACHE	MILD	N
01-002	NAUSEA	SEVERE	Y

10. Key Points and Best Practices

Always create new variables for recoded values to preserve the original SDTM data.
Use vectorized functions (lapply, mutate(across())) for efficiency.
For categorical/factor variables, use factor() or forcats for level recoding.
Document your recoding logic for traceability and regulatory compliance.
For complex recoding, custom functions or lookup tables can be helpful.

Resource download links

1.4.15.-Recoding-Variables-in-SAS-vs-R.zip

⁂

1.4. Migrating from SAS to R: A Skill Conversion Guide