Preoperative sizing of hip hemiarthroplasties to accurately estimate head size from non- standardised pelvic radiographs: Can it be done?

Background: Preoperative sizing of implants for hip fracture patients requiring a hemiarthroplasty is difficult due to non-standardised radiographs, absence of sizing marker, variable patient position and body habitus. We investigated whether a simple tool could help predict femoral head size, allowing surgeons to safely proceed with surgery when implant stocks are limited, and to potentially improve theatre efficiency.

Methods: Three independent reviewers measured the maximum width of the contralateral (intact) femoral head using PACS software in 50 cases of intracapsular hip fracture. This was linearly regressed on actual implant size to calculate the average magnification coefficient. Inter- and intra-rater reliability were evaluated using intraclass correlation coefficients (ICC).

Results: The best fitting magnification constant was 118% (95% confidence interval 16.0-19.7%), which achieved a mean error of 1.7 mm. Prediction accuracy was significantly improved by allowing a constant (intercept) as a second parameter in the regression model (p = 0.01), which achieved a mean error of just 1.4 mm from the implant used. The inclusion of the constant reduces errors at the upper and lower extremes of head sizes. ICCs for inter- and intra-rather agreement were 0.94 and 0.98 respectively.

Conclusion(s): We have shown that hip hemiarthroplasty head sizes can be reliably and accurately predicted from non-standardised pre-operative radiographs. We have devised a method which can easily be adopted by other centres and tailored to the characteristics of their radiology department.