diff --git a/src/nested_pandas/series/dtype.py b/src/nested_pandas/series/dtype.py
index 3559b7b..7ae5302 100644
--- a/src/nested_pandas/series/dtype.py
+++ b/src/nested_pandas/series/dtype.py
@@ -13,7 +13,6 @@
 from pandas.core.arrays import ExtensionArray
 from pandas.core.dtypes.base import ExtensionDtype
 
-from nested_pandas.series.na import NA, NAType
 from nested_pandas.series.utils import is_pa_type_a_list
 
 __all__ = ["NestedDtype"]
@@ -29,9 +28,9 @@ class NestedDtype(ExtensionDtype):
     """Attributes to use as metadata for __eq__ and __hash__"""
 
     @property
-    def na_value(self) -> NAType:
+    def na_value(self) -> pd.NA:
         """The missing value for this dtype"""
-        return NA
+        return pd.NA
 
     type = pd.DataFrame
     """The type of the array's elements, always pd.DataFrame"""
diff --git a/src/nested_pandas/series/ext_array.py b/src/nested_pandas/series/ext_array.py
index df0b99f..efce5f0 100644
--- a/src/nested_pandas/series/ext_array.py
+++ b/src/nested_pandas/series/ext_array.py
@@ -160,8 +160,16 @@ def to_numpy(self, dtype: None = None, copy: bool = False, na_value: Any = no_de
 
         # Hack with np.empty is the only way to force numpy to create 1-d array of objects
         result = np.empty(shape=array.shape, dtype=object)
+
+        # TODO: delete block once confirmed NA is better to return
+        # fill NaNs - Fill with an empty dictionary using the same columns
+        # nan_vals = pd.isna(array)
+        # if sum(nan_vals) > 0:
+        #    na_idx = np.where(nan_vals)
+        #    array[na_idx] = {key: [] for key in array[np.where(~nan_vals)[0][0]].keys()}
+
         # We do copy=False here because user's 'copy' is already handled by ArrowExtensionArray.to_numpy
-        result[:] = [pd.DataFrame(value, copy=False) for value in array]
+        result[:] = [pd.DataFrame(value, copy=False) if not pd.isna(value) else pd.NA for value in array]
         return result
 
     def __setitem__(self, key, value) -> None:
diff --git a/src/nested_pandas/series/na.py b/src/nested_pandas/series/na.py
deleted file mode 100644
index 0b77bb4..0000000
--- a/src/nested_pandas/series/na.py
+++ /dev/null
@@ -1,55 +0,0 @@
-"""Missing value for NestedDtype
-
-It i something between pandas' NA and NaN
-"""
-
-__all__ = ["NAType", "NA"]
-
-
-class _NAType:
-    pass
-
-
-class NAType:
-    """Singleton class representing missing value for NestedDtype.
-
-    It doesn't implement most of the arithmetics and boolean logic operations,
-    because they are ambiguous for missing values.
-
-    The implementation is inspired both by pandas' NA and float number NaN.
-
-    `NA` is a singleton instance of this class.
-    """
-
-    _instance = None
-
-    def __new__(cls, *args, **kwargs):
-        """Create a new instance of NAType."""
-        if cls._instance is None:
-            cls._instance = super().__new__(cls)
-        return cls._instance
-
-    def __repr__(self) -> str:
-        return "<NA>"
-
-    def __format__(self, format_spec) -> str:
-        try:
-            return self.__repr__().__format__(format_spec)
-        except ValueError:
-            return self.__repr__()
-
-    def __bool__(self):
-        raise TypeError("boolean value of NA is ambiguous")
-
-    def __eq__(self, other):
-        return False
-
-    def __ne__(self, other):
-        return True
-
-    def __hash__(self):
-        return 0
-
-
-NA = NAType()
-"""Missed value for NestedDtype, a singleton instance of `NAType` class."""
diff --git a/tests/nested_pandas/series/test_dtype.py b/tests/nested_pandas/series/test_dtype.py
index 97aa32d..2b6b3b2 100644
--- a/tests/nested_pandas/series/test_dtype.py
+++ b/tests/nested_pandas/series/test_dtype.py
@@ -3,7 +3,6 @@
 import pytest
 from nested_pandas.series.dtype import NestedDtype
 from nested_pandas.series.ext_array import NestedExtensionArray
-from nested_pandas.series.na import NA
 
 
 @pytest.mark.parametrize(
@@ -62,7 +61,7 @@ def test_from_fields():
 def test_na_value():
     """Test that NestedDtype.na_value is a singleton instance of NAType."""
     dtype = NestedDtype(pa.struct([pa.field("a", pa.list_(pa.int64()))]))
-    assert dtype.na_value is NA
+    assert dtype.na_value is pd.NA
 
 
 def test_fields():
diff --git a/tests/nested_pandas/series/test_na.py b/tests/nested_pandas/series/test_na.py
deleted file mode 100644
index b3a81b3..0000000
--- a/tests/nested_pandas/series/test_na.py
+++ /dev/null
@@ -1,50 +0,0 @@
-import pytest
-from nested_pandas.series.na import NA
-
-
-def test_na_is_singleton():
-    """Test that NA is a singleton instance"""
-    assert NA is NA
-
-
-def test_na_repr():
-    """Test that NA has the correct representation."""
-    assert repr(NA) == "<NA>"
-
-
-def test_na_format():
-    """Test that NA has the correct format."""
-    assert f"{NA}" == "<NA>"
-
-
-def test_na_bool():
-    """Test that NA raises TypeError when converted to bool."""
-    with pytest.raises(TypeError):
-        bool(NA)
-
-
-def test_na_eq():
-    """Test that NA is not equal to anything."""
-    assert NA != 1
-    assert NA != 1.0
-    assert NA != "1"
-    assert NA != NA
-
-
-def test_na_neq():
-    """Test that NA is not equal to anything."""
-    assert NA != 1
-    assert NA != 1.0
-    assert NA != "1"
-    assert [] != NA
-    assert {} != NA
-    assert NA != ()
-    assert set() != NA
-    assert NA != NA
-    assert object() != NA
-
-
-def test_hash():
-    """Test that hash(NA) is always the same."""
-    assert hash(NA) == hash(NA)
-    assert {NA, NA} == {NA}