Determination of suitable sample sizes for multi-patient based finite element studies

Finite element analysis is used extensively to assess joint replacements, but the majority of these are single sample studies. Recent investigations have suggested that such studies are unable to account for natural inter-patient variation in bone geometry and material property distribution. Recent developments in computer tomography based analyses make multiple sample studies possible; the question remains how many femurs are required to perform a study which accounts for such variations. This work investigates the factors that should be considered in answering this question. It explores sample sizing techniques when comparing strain distribution in the intact and implanted femur and when comparing two or more implant designs in a group of femurs. An example analysis of the effect of femoral head resurfacing was undertaken. Two sample sizing calculations were utilised, one based on achieving the desired precision in results, the other based on determining if a significant difference exists between two designs. The analysis shows that reasonable statistical precision can be achieved with a group of femurs. The study was also able to determine a suitable sample size for the analysis of a statistically significant difference between two groups of femurs with varying design parameters. The study concluded that while sample sizing is recommended for an accurate analysis, consideration must be made for the practicality of such a task.