The authors are to be congratulated on their extensive case-control study of the sagittal geometry of the lateral knee compartment and its association with non-contact anterior cruciate ligament (ACL) injury. The underlying motivation for identifying risk factors is to recognize individuals at high risk for ACL injury and to intervene (when possible) to reduce the risk in those individuals. Unfortunately, if the results and conclusions of the study are reviewed from the perspective of a clinician hoping to identify individuals at high risk for non-contact ACL injury, there is no “tool” or guideline for how the risk factors can be used to assist the clinician.
Female athletes have a risk of non-contact ACL injury that is two to five times that of male athletes. One might therefore expect female athletes to demonstrate the greatest effect size in the risk factors associated with non-contact ACL injury. However, the results of this study show no difference between injured and uninjured female subjects with respect to any of the six geometric parameters pertaining to the lateral knee compartment—the maximal anteroposterior femoral articular width (FAP), distal femoral radius of curvature (Fr), maximal anteroposterior tibial articular width (TPAP), tibial plateau radius of curvature (TPr), ratio of femoral width to tibial width (FAP:TPAP), or ratio of femoral to tibial radius of curvature (Fr:TPr). Indeed, the corresponding p values (0.65, 0.83, 0.77, 0.68, 0.75, and 0.80) do not even approach significance. What value for any of these measured parameters would a clinician use as a cutoff to identify a high-risk female athlete?
The authors report thirty-six statistical comparisons in six tables. If a corresponding Bonferroni correction is applied to the significance level, the p value corresponding to a statistically significant difference would change from 0.05 to 0.05/36 = 0.0014. Fifteen of the thirty-six comparisons are significant at the corrected significance level, and five of these significant differences are between the uninjured female and uninjured male subgroups. The authors need to specify a priori comparisons to specifically address their hypothesis and not simply perform all possible comparisons between the male and female, injured and uninjured, and combined and uncombined subgroups. How do significant differences between uninjured female and uninjured male athletes support their hypothesis?
Ideal screening methods for risk factors need to be relatively low-cost and simple, with minimal time requirements. Use of the method in this paper would require prospective magnetic resonance (MR) imaging of all athletes as well as relatively complex digital analysis to measure the parameters. Furthermore, these parameters would be dynamic during skeletal maturation. At what age would MR imaging be (ideally) performed?
The authors report “almost perfect” interobserver agreement with their method of MR image measurement. However, they reduced the intraobserver and interobserver variability with two steps. First, “each series of T2-weighted proton-density MRIs was prescreened by the senior author to ensure that the axial, coronal, and sagittal reconstructed images were orthogonal to the posterior femoral condyles.” Second, “this sagittal reference plane was agreed on by the three observers and was used for all measurements in this study.” Ideally, each observer would have been given three-dimensional MR images and would have independently determined the planes orthogonal to the posterior femoral condyles, independently determined the femoral longitudinal axis, independently determined the sagittal reference plane, and then independently measured the sagittal-plane femoral and tibial parameters. Each observer would then have repeated these measurements for the same subject multiple times to calculate intraobserver as well as interobserver correlation coefficients. This four-step process would likely have yielded much lower values for these correlation coefficients.
The authors have articulated a set of possible sagittal-plane anatomic risk factors for a non-contact ACL injury. However, they have not demonstrated any correlation involving these risk factors in female subjects, arguably the highest-risk population. No cutoff values for identifying high-risk athletes are articulated. The reproducibility of their method also requires further quantification to judge its value.
Furthermore, use of these measurements as a screening tool for high-risk athletes would require obtaining MR images in this at-risk population; this is a formidable task. Perhaps this method could find clinical utility as a screening tool to help counsel patients who are contemplating a revision ACL reconstruction regarding their risk of reinjury. For the method to be applicable, it would need to be applied to standard MR imaging sequences, with good reproducibility and better-defined parameters for the predictive ability of this set of anatomic variables.
Even without clinical utility, the authors have further refined a particular set of anatomic variables that may play an important role in the risk equation for non-contact ACL injury.