We have compared the changes in the pattern of the principal strains in the proximal femur after insertion of eight uncemented anatomical stems and eight customised stems in human cadaver femora. During testing we aimed to reproduce the physiological loads on the proximal femur and to simulate single-leg stance and stair-climbing.

The strains in the intact femora were measured and there were no significant differences in principal tensile and compressive strains in the left and right femora of each pair. The two types of femoral stem were then inserted randomly into the left or right femora and the cortical strains were again measured. Both induced significant stress shielding in the proximal part of the metaphysis, but the deviation from the physiological strains was most pronounced after insertion of the anatomical stems.

The principal compressive strain at the calcar was reduced by 90% for the anatomical stems and 67% for the customised stems. Medially, at the level of the lesser trochanter, the corresponding figures were 59% and 21%. The anatomical stems induced more stress concentration on the anterior aspect of the femur than did the customised stems. They also increased the hoop strains in the proximomedial femur. Our study shows a consistently more physiological pattern of strain in the proximal femur after insertion of customised stems compared with standard, anatomical stems.

CT and advanced computer-aided design techniques offer the means for designing customised femoral stems. Our aim was to determine the Hounsfield (HU) value of the bone at the corticocancellous interface, as part of the criteria for the design algorithm.

We obtained transverse CT images from eight human cadaver femora. The proximal femoral canal was rasped until contact with dense cortical bone was achieved. The femora were cut into several sections corresponding to the slice positions of the CT images. After obtaining a computerised image of the anatomical sections using a scanner, the inner cortical contour was outlined and transferred to the corresponding CT image. The pixels beneath this contour represent the CT density of the bone remaining after surgical rasping. Contours were generated automatically at nine HU levels from 300 to 1100 and the mean distance between the transferred contour and each of the HU-generated contours was computed.

The contour generated along the 600-HU pixels was closest to the inner cortical contour of the rasped femur and therefore 600 HU seem to be the CT density of the corticocancellous interface in the proximal part of cadaver femora. Generally, femoral bone with a CT density beyond 600 HU is not removable by conventional reamers. Thus, we recommend the 600 HU threshold as one of several criteria for the design of custom femoral implants from CT data.