Progressive retroversion of a cemented stem is predictive of early loosening and failure. We assessed the relationship between direct post-operative stem anteversion, measured with CT, and the resulting rotational stability, measured with repeated radiostereometric analysis over ten years. The study comprised 60 cemented total hip replacements using one of two types of matt collared stem with a rounded cross-section. The patients were divided into three groups depending on their measured post-operative anteversion (< 10°, 10° to 25°, > 25°). There was a strong correlation between direct post-operative anteversion and later posterior rotation. At one year the < 10° group showed significantly more progressive retroversion together with distal migration, and this persisted to the ten-year follow-up. In the < 10° group four of ten stems (40%) had been revised at ten years, and an additional two stems (20%) were radiologically loose. In the 'normal' (10° to 25°) anteversion group there was one revised (3%) and one loose stem (3%) of a total of 30 stems, and in the > 25° group one stem (5%) was revised and another loose (5%) out of 20 stems. This poor outcome is partly dependent on the design of this prosthesis, but the results strongly suggest that the initial rotational position of cemented stems during surgery affects the subsequent progressive retroversion, subsidence and eventual loosening. The degree of retroversion may be sensitive to prosthetic design and stem size, but < 10° of anteversion appears deleterious to the long-term outcome for cemented hip prosthetic stems.