diff --git a/NEWS.md b/NEWS.md
index 3330cf8cb..2c60c7193 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -1,5 +1,7 @@
 # chevron 0.2.8.9006
 
+* `ttet01_main` now prints a clearer error message when a level in `arm_var` variable is missing.
+
 # chevron 0.2.8
 
 * New `unwrap` argument prints `preprocessing`, `main`, `postprocessing` and `layout` function upon execution of the `run` method.
diff --git a/R/ttet01.R b/R/ttet01.R
index 6f14caf18..5594f8a68 100644
--- a/R/ttet01.R
+++ b/R/ttet01.R
@@ -46,9 +46,10 @@ ttet01_main <- function(adam_db,
   assert_valid_variable(adam_db[[dataset]], c("IS_EVENT", "IS_NOT_EVENT"), types = list("logical"), label = df_label)
   assert_valid_variable(adam_db[[dataset]], "AVAL", types = list("numeric"), lower = 0, label = df_label)
   assert_valid_variable(
-    adam_db[[dataset]], c("USUBJID", strata, arm_var, "EVNT1", "EVNTDESC", "AVALU"),
+    adam_db[[dataset]], c("USUBJID", strata, "EVNT1", "EVNTDESC", "AVALU"),
     types = list(c("character", "factor")), label = df_label
   )
+  assert_valid_variable(adam_db[[dataset]], arm_var, types = list("factor"), empty.levels.ok = FALSE, label = df_label)
   assert_subset(ref_group, lvls(adam_db[[dataset]][[arm_var]]))
   ref_group <- ref_group %||% lvls(anl[[arm_var]])[1]
   assert_single_value(anl$AVALU, label = sprintf("adam_db$%s$AVALU", dataset))
diff --git a/tests/testthat/test-ttet01.R b/tests/testthat/test-ttet01.R
index 609ad9e02..6c617983d 100644
--- a/tests/testthat/test-ttet01.R
+++ b/tests/testthat/test-ttet01.R
@@ -10,6 +10,18 @@ test_that("ttet01 works as expected", {
   expect_snapshot(cat(export_as_txt(res, lpp = 100)))
 })
 
+test_that("ttet01 returns an error when one level or arm_var is empty", {
+  proc_data <- syn_data %>%
+    dunlin::log_filter(PARAMCD == "PFS", "adtte") %>%
+    dunlin::log_filter(ARM != "C: Combination", "adtte")
+
+  expect_error(
+    run(ttet01, proc_data, dataset = "adtte"),
+    "* Variable 'adam_db$adtte$ARM': Has has empty levels 'C: Combination'.",
+    fixed = TRUE
+  )
+})
+
 test_that("ttet01 works as expected for stratified and unstratified analysis", {
   skip_on_os("windows")
   filter_data <- dunlin::log_filter(syn_data, PARAMCD == "PFS", "adtte")