Hello Patryk,

I see the threads on the screw. Was that screw solid body subtracted from all the bone materials to make a screw-shaped hole in the bone?  I assume that is what you have done, but I ask just to make sure.

Cortical bone is a brittle material, it fractures, it does not deform the way a ductile metal would plastically deform. Therefore you should not use any material models from the plasticity section for cortical bone. Cancellous bone, since it is full of voids may behave differently.

I assume metals used for medical screws have high strength, so when they pull out, it is the bone that fails, not the metal.

What data do you know for the material properties of each type of bone? I assume you know at least the Young’s modulus (E) and Ultimate Tensile Strength (UTS). 

Cortical bone is much stronger than cancellous bone.  The first few threads carry the majority of the stress in a screw and those are in the cortical bone, so when that fails, the screw pulls out. The contribution of the cancellous bone may be insignificant.

You need a material failure model at least for the cortical bone. Explicit Dynamics has a fully automated means of deleting elements that have failed. The simplest method is to remove elements that have exceeded a failure strain, which could be simply computed as UTS/E but more sophisticated material failure models can be substituted. With Explicit Dynamics, you have to build a mesh that avoids elements with tiny element edge lengths to prevent the automatically calculated time step from becoming very, very small, which causes the computation time to become very, very long.

If you stay with implicit solvers such as Static Structural, there is a command called EKILL that can remove the stiffness from failed elements during the solution, but this requires you to use APDL code in your model to remove failed elements after each load increment. There are some threads on this topic you can read.