Background: In a previous randomized studiy using Röntgen Stereometric Analysis (RSA), we showed that oral bisphosphonates reduce the mean migration distance during the first 6 months. In a similar randomized study, bisphosphonates applied locally at the operation had a similar effect. These studies showed a 0.1 mm difference in mean value between groups. Does such a small difference matter? We addressed this question by use of frequency analysis.

Methods: The 2 previous studies were combined for analysis, and designated as bisphosphonate (n=44) or control treated (n=49). We analysed the migration vector (for the center of the rigid body) by use of a set of algoritms for frequency distribution analysis called Rmix. The migration vector lengths were assumed to be a compound of log-normal distributions. The frequency analysis determined if the observed frequency distributions were best described as a single, or a sum of 2 or more lognormally distributed subgroups.

Results: After 6 months, the control patients had formed 2 subgroups, one comprising 85% of the patients. The dichotomy was significant (p=0.016).

After 2 years, the dichotomy persisted (p=0.027). In the bisphosphonate-treated patients, no dichotomies could be found. The distribution of the migration vector length appeared similar to the larger and less migrating subgroup among the controls.

Discussion: The risk of aseptic loosening for cemented knees is extremely small. However, the migrating subgroup among our control patients may be at risk of loosening, and would have run a high risk if they were young and active. This subgroup did not appear with bisphosphonate treatment

Summary: In previous comparisons we found a slight decrease in mean value with bisphosphonates. The present analysis shows that this reflects the disappearance of a small subgroup with large migration.