We present a comparative Density Functional Theory (DFT) study based on two different implementations of relativistic effects within the Kohn-Sham (KS) approach, to describe the metal-ligand interaction in I(3)M-L complexes (L = NH(3), NCCH(3), CO and M = La, Nd, U). In the first model, the scalar corrections were included by a quasi-relativistic approach (QR) via the so-called ZORA or Pauli Hamiltonians, while in the second, these effects are taken into account in a quasi-Relativistic Effective Core Potential (RECP). These relativistic approaches were used in conjunction with various gradient corrected (GGA) or hybrid (SCH) functionals. The structural parameters obtained from geometry optimizations have been compared to experimental structural trends, and rationalized by a KS orbital analysis. Both approaches provide similar results for mainly ionic metal-ligand bonds (e.g., for the sigma-donor ligand L = NH(3)). For the pi-acceptor ligands (NCCH(3), CO), the QR approach is in agreement with experimental trends and consistent with the presence of a backbonding interaction between U(III) and the neutral ligand, which does not exist in the lanthanide homologues. The GGA/RECP methods also reproduce this phenomenon, while the SCH/RECP scheme fails to describe this interaction. The role of the RECP, of its size, and of additional polarization functions has also been examined. Finally, the failure of the SCH/RECP approach was interpreted as a consequence of a bad estimation of frontier orbital energy levels in the uranium and ligand species.