In "Knots and Links in Spatial Graphs" (Journal of Graph Theory, vol.7 1983, 445-453), Conway and Gordon write :
"Now it is a standard fact in knot theory, not hard to prove, that any two spatial embeddings of $\Gamma$ are equivalent under the equivalence relation generated by moves of the form: ambiant isotopy to regular projection position followed by a change of crossing from over to under (a crossing change)."
Here $\Gamma$ is a finite graph, and a spatial embedding is a tame (polygonal) embedding into Euclidean space.
I am not sure of the exact meaning of this sentence, and would like to have some references where this would be stated and proved.