In a previous preclinical study the prototype version of a partially cemented hip stem, cement-locked uncemented (CLU) prosthesis, showed optimal primary stability and moderate stress shielding. However, numerical analysis suggested that the prototype design would induce relatively high stresses in the cement and a significant relative motion between cement and metal. The present study aimed to verify if these problems could be eliminated once the CLU design is improved. The revised design was analysed using a complete finite element model of an implanted human femur. To further strengthen the predictions of the finite element analysis, the cement damage induced by a severe load history was assessed experimentally in synthetic femurs implanted with the improved CLU stem or with a clinically successful fully cemented stem. The modifications made to the CLU stem design did not reduce its good primary stability but decreased the metal-cement relative micromotion. The same load induced stresses in the cement mantle of the improved CLU stem that were significantly lower than those predicted for the prototype design. Although the presence of modelling artefacts produced a highly localized stress peak of 13 MPa. 99 per cent of the cement volume was subjected to a principal tensile stress lower then 4 MPa. These levels of stress compare favourably with the tensile fatigue limit of the acrylic cement used in this study (9.7 MPa). The experimental results further supported these findings. The cemented stem showed a number of cracks per volume unit approximately ten times higher than the partially cemented stem under investigation.