A new method to measure anatomic knee alignment for large studies of OA: data from the osteoarthritis initiative

To develop and validate a new and improved software method to rapidly determine femur-tibia angle (FTA).

Three readers, two skilled and one unskilled, without any formal medical training, measured FTA in 142 subjects from the Osteoarthritis Initiative (OAI). The reader reliability was assessed using the intra-class correlation coefficient (ICC), root mean square standard deviation (RMSSD), and Bland-Altman plots, comparing the existing and new FTA methods. Gender-specific linear regression assessed the relationship of FTA with the hip-knee-ankle angle (HKA).

The ICC (RMSSD) for intra- and inter-reader reproducibility of the existing FTA method was 0.96 (0.77°) and 0.92 (1.38°), respectively, and for the new technique was 0.98 (0.25°) and 0.98 (0.37°), with similar results for all three readers. Bland-Altman 95% limits of agreement were greater than ±2° for the existing, and ±1° for the new method. The r-value for the relation of FTA to HKA was 0.68 and 0.72 for the existing and new methods, respectively. Varus (HKA ≤ -2°)/neutral (-2° < HKA < 2°)/valgus (HKA ≥ 2°) alignment based on predicted HKA agreed moderately with measured HKA (weighted kappa = 0.53), and had moderate sensitivity (73%) and specificity (84%) for varus malalignment. The new FTA was related to HKA using a linear equation with a slope of 0.98 and an offset of 4.0°.

Since it is largely automated and uses unambiguous anatomical landmarks, the new method is highly reproducible and can be made on a standard posteroanterior (PA) knee radiograph by a relatively unskilled reader.