diff --git a/NEWS.md b/NEWS.md
index ec1d0d80..f147140a 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -99,6 +99,7 @@ the dosing including dose amount and route.
 when the issue is due to an excluded point (#310)
 * The `PKNCAdose` function won't give an error for a missing-time check when the issue is due to an excluded point (#310)
 * `pk.nca` will calculate `fe` and `clr` even if their dependent parameters (e.g, `ae`) were not requested to be calculated in the intervals (#473)
+* sparse calculations won't abort with `pk.nca` when the data contains missing (NA) concentrations. It will silently drop them.
 
 ## New features
 
diff --git a/R/sparse.R b/R/sparse.R
index dcbb32f6..6bac6f92 100644
--- a/R/sparse.R
+++ b/R/sparse.R
@@ -14,8 +14,15 @@ as_sparse_pk <- function(conc, time, subject) {
     subject <- conc$subject
     conc <- conc$conc
   }
-  assert_conc_time(conc = conc, time = time, any_missing_conc = FALSE, sorted_time = FALSE)
+  assert_conc_time(conc = conc, time = time, any_missing_conc = TRUE, sorted_time = FALSE)
   checkmate::check_vector(subject, any.missing=FALSE, len=length(conc), null.ok=FALSE)
+  # Drop observations with missing concentrations so that per-timepoint means,
+  # variances, and subject counts reflect only available data.
+  mask_ok <- !is.na(conc)
+  conc <- conc[mask_ok]
+  time <- time[mask_ok]
+  subject <- subject[mask_ok]
+
   unique_times <- sort(unique(time))
   ret <- list()
   for (current_time in unique_times) {
@@ -632,4 +639,4 @@ PKNCA.set.summary(
   description = "arithmetic mean and standard deviation",
   point = business.mean,
   spread = business.sd
-)
\ No newline at end of file
+)
diff --git a/tests/testthat/test-sparse.R b/tests/testthat/test-sparse.R
index e685d09e..98f7cf9d 100644
--- a/tests/testthat/test-sparse.R
+++ b/tests/testthat/test-sparse.R
@@ -33,6 +33,30 @@ test_that("sparse_auc", {
   expect_equal(sparse_serial$sparse_auc_df, structure(auclast_df_serial, method=c("AUC: linear", "Sparse: arithmetic mean, <=50% BLQ")))
 })
 
+test_that("as_sparse_pk drops NA concentrations (#563)", {
+  sparse_pk <- as_sparse_pk(
+    conc    = c(0, 0, 5, 7, 3, NA),
+    time    = c(0, 0, 4, 4, 24, 24),
+    subject = c(1, 2, 3, 4, 5, 6)
+  )
+  # The NA row (subject 6, time 24) should be dropped
+  expect_length(sparse_pk, 3)  # 3 unique times: 0, 4, 24
+  expect_equal(sparse_pk[[3]]$conc, 3)
+  expect_equal(sparse_pk[[3]]$subject, 5)
+})
+
+test_that("sparse_auc tolerates NA concentrations (#563)", {
+  result <- pk.calc.sparse_auc(
+    conc    = c(0, 0, 5, 7, 3, NA),
+    time    = c(0, 0, 4, 4, 24, 24),
+    subject = 1:6
+  )
+  expect_false(is.na(result$sparse_auc))
+  # Mean profile: time 0 = 0, time 4 = 6, time 24 = 3
+  # AUC linear: 0.5*(0+6)*4 + 0.5*(6+3)*20 = 12 + 90 = 102
+  expect_equal(as.numeric(result$sparse_auc), 102)
+})
+
 test_that("sparse_auclast expected errors", {
   expect_error(
     pk.calc.sparse_auclast(auc.type = "foo"),