← SASnRWrite SAS Transport Files (.xpt)

Introduction

FDA regulatory submissions require datasets in SAS V5 transport format (.xpt) — this is often the very last step in a clinical programming workflow.
Both SAS and R can produce compliant .xpt files, so the choice of environment does not affect the submission.
Key constraints to check before writing: variable names must be 8 characters or fewer, and variable labels must be 200 characters or fewer.
SAS uses LIBNAME XPORT with PROC COPY; R uses haven::write_xpt() with version = 5.
After writing, we verify the .xpt file by reading it back in and confirming that labels, types and row counts are intact.

Write SAS Transport Files (.xpt)

SAS (Base SAS)+

*==============================================================================;
* Write SAS Transport Files (.xpt)
*==============================================================================;

*------------------------------------------------------------------------------;
* Step 1 - Create a small ADSL dataset with labels
*------------------------------------------------------------------------------;
data adsl;
    length USUBJID $20 SITEID $4 SEX $1 TRT01P $20;
    label USUBJID = "Unique Subject Identifier"
          SITEID  = "Study Site Identifier"
          SEX     = "Sex"
          AGE     = "Age in Years"
          TRT01P  = "Planned Treatment for Period 01";
    infile datalines dlm='|' dsd missover;
    input USUBJID : $20. SITEID : $4. SEX : $1. AGE : best32. TRT01P : $20.;
    datalines;
ABC-001-001|001|M|45|Placebo
ABC-001-002|001|F|52|Drug A
ABC-002-003|002|M|38|Drug B
ABC-002-004|002|F|61|Placebo
ABC-003-005|003|M|55|Drug A
;
run;

*------------------------------------------------------------------------------;
* Step 2 - Write the .xpt file using LIBNAME XPORT + PROC COPY
*------------------------------------------------------------------------------;
libname xptout xport "C:\clinical\output\adsl.xpt";

proc copy in=work out=xptout;
    select adsl;
run;

libname xptout clear;

*------------------------------------------------------------------------------;
* Step 3 - Verify by reading it back
*------------------------------------------------------------------------------;
libname xptin xport "C:\clinical\output\adsl.xpt";

proc contents data=xptin.adsl;
run;

proc print data=xptin.adsl;
run;

libname xptin clear;

*------------------------------------------------------------------------------;
* Alternative - PROC CPORT (creates a different binary format)
*------------------------------------------------------------------------------;
* proc cport data=adsl file="C:\clinical\output\adsl_cport.xpt";
* run;
* Note: PROC CPORT creates a SAS transport file that only SAS can read.
* For FDA submissions, use the LIBNAME XPORT method above.;

We first build a small ADSL dataset with labels assigned, since labels are preserved in the transport file and required for submissions.
To write the .xpt file, we assign a LIBNAME with the XPORT engine pointing to the output file path.
PROC COPY transfers the dataset from WORK to the XPORT library — SELECT adsl; tells SAS which dataset to write.
We clear the libname after writing with LIBNAME xptout CLEAR; to release the file handle.
To verify the output, we assign a new read-only libname to the .xpt file and run PROC CONTENTS and PROC PRINT.
PROC CPORT is an alternative, but it creates a proprietary format that only SAS can read — FDA reviewers expect the standard XPORT format.
Before writing, we confirm that all variable names are 8 characters or fewer and all labels are 200 characters or fewer — the V5 transport format enforces these limits.

R (tidyverse)+

#==============================================================================
# Write SAS Transport Files (.xpt)
#==============================================================================

library(tidyverse)
library(haven)
library(labelled)

#------------------------------------------------------------------------------
# Step 1 - Create a small ADSL dataset
#------------------------------------------------------------------------------
adsl <- tribble(
  ~USUBJID,      ~SITEID, ~SEX, ~AGE, ~TRT01P,
  "ABC-001-001", "001",   "M",  45,   "Placebo",
  "ABC-001-002", "001",   "F",  52,   "Drug A",
  "ABC-002-003", "002",   "M",  38,   "Drug B",
  "ABC-002-004", "002",   "F",  61,   "Placebo",
  "ABC-003-005", "003",   "M",  55,   "Drug A"
)

#------------------------------------------------------------------------------
# Step 2 - Assign variable labels
#------------------------------------------------------------------------------
adsl <- adsl %>%
  set_variable_labels(
    USUBJID = "Unique Subject Identifier",
    SITEID  = "Study Site Identifier",
    SEX     = "Sex",
    AGE     = "Age in Years",
    TRT01P  = "Planned Treatment for Period 01"
  )

#------------------------------------------------------------------------------
# Step 3 - Check variable name lengths (must be <= 8 for V5)
#------------------------------------------------------------------------------
long_names <- names(adsl)[nchar(names(adsl)) > 8]
if (length(long_names) > 0) {
  warning("Variables exceed 8-char limit: ", paste(long_names, collapse = ", "))
}

#------------------------------------------------------------------------------
# Step 4 - Write the .xpt file (version 5 for FDA)
#------------------------------------------------------------------------------
write_xpt(adsl, "C:/clinical/output/adsl.xpt", version = 5)

#------------------------------------------------------------------------------
# Step 5 - Verify by reading it back
#------------------------------------------------------------------------------
adsl_check <- read_xpt("C:/clinical/output/adsl.xpt")
glimpse(adsl_check)

adsl_check %>%
  map_chr(~ attr(.x, "label") %||% "(no label)") %>%
  enframe(name = "variable", value = "label")

We load haven for the transport file I/O and labelled for the convenient set_variable_labels() function.
The sample dataset is built with tribble() using the same five subjects and labels as the SAS version.
set_variable_labels() from the labelled package assigns labels in a single piped call — each argument is VARNAME = "Label text".
Before writing, we check that no variable name exceeds 8 characters — the V5 transport format silently truncates long names, which can break downstream programs.
write_xpt(adsl, path, version = 5) writes the FDA-compliant transport file. The version = 5 argument is required — the default is version 8, which some FDA tools cannot read.
We read the file back with read_xpt() and run glimpse() to confirm row counts, column types and labels survived the round trip.
The label check at the end prints every variable and its label so we can visually compare against the original assignments.

Happy learning!

Search lessons

Introduction

Write SAS Transport Files (.xpt)