Abstract The factors responsible for clinical stem failures of total hip joint prostheses were studied by instrumenting the Charnley, Mueller and Trapezoidal-28 prostheses with strain gauges and subjecting them to various conditions of load and acrylic fixation. Loading tests of stems which were supported as simple cantilever beams demonstrated that the Mueller prosthesis was comparatively weak in regions of observed clinical stem fractures. Simulated femoral loading tests of prostheses securely fixed in acrylic bone cement showed low stress levels in the stems and necks. Varus placement of the stems caused a marked increased stress in the stem; this condition should be avoided at surgery. Our tests showed that a prosthesis which is loose in the acrylic also raises stress levels in the distal stem. As acrylic is setting, movement of the prosthesis must be avoided. Lack of adequate proximal medial support for the stem, due to poor quality bone stock or acrylic fracture, will create dangerously high stress levels in the stem: this emphasizes the importance of good cement filling technique and proper patient selection. Three clinical stem failures have been analyzed using the results of these tests. We have concluded that inadequate stem design and varus placement were contributing factors in the fatigue fracture of a cast cobalt-chromium-molybdenum Mueller stem. Contributing factors of the failure of two cast 316 LVM stainless steel Bechtol stems were low strength metal and technical error in cementing one of them which was apparently loose in the acrylic at the time of surgery.