On the basis of the Norwegian Arthroplasty Register, which has recorded nearly all primary hip prostheses and revisions in Norway since 1987, we studied risk factors for prosthesis luxation leading to revision. 7 prosthesis brand combinations used in 42,987 primary operations were included from 1987-2000. We found that femoral head size was an important risk factor; 28 mm heads led to revision more often than 32 mm ones (failure rate ratio (FRR) 4.0, 95% confidence interval (CI) 2.2-7.3). Charnley (22 mm head) performed equally well or better than the 28 mm heads. The Exeter stem and cup is the type of prosthesis on the Norwegian market with more than two femoral head sizes (26, 28, 30, 32 mm) and 26 mm heads led to revision due to luxation significantly more often than 30 mm heads (FRR 4.1, 95%CI 2.2-8.1). Old age, preoperative diagnosis, and choice of prosthesis brand combination were also important factors affecting the revision rate due to luxation. A posterior approach increased the risk of revision more than a lateral one (FRR 1.9, 95% CI 1.4-2.5). Gender, trochanteric osteotomy and duration of the operation did not affect the results.