Summary

We have presented a modified, practical method to predict the ACL hamstring graft size with high specificity using pre-operative MRI measurements that does not require any specialized software or methods, and can be reliably done even by junior members of the surgical team.

Abstract

Background

Predicting hamstring graft size pre-operatively for the surgical reconstruction of the anterior cruciate ligament (ACL) is important as it may help pre-empt an insufficient diameter in graft size intra-operatively, which may lead to graft failure. While there are multiple published models for the prediction of the hamstring graft pre-operatively using magnetic resonance imaging (MRI) picture archiving and communication systems (PACS), most are not feasible and practical. We aim to ask the following questions: (1) Can an ACL hamstring graft size be practically predicted in a numerically continuous manner, by surgeons or surgical assistants of all levels of training, using the pre-operative MRI from any native MRI PACS system? (2) Using this method of prediction, what is the degree of correlation between the predicted and actual graft size? (3) If we define an adequate actual graft size as more than or equal to 8mm, what is the performance of this method of prediction in terms of specificity, sensitivity and discriminative ability?

Methods

A retrospective review of 112 patients who underwent primary ACL reconstruction with quadrupled hamstring semitendinosus-gracilis grafts at a tertiary institution between January 2018 and December 2018 was conducted. Two independent and blinded evaluators with no prior radiology posting experience measured the cross-sectional lengths and breadths of both semitendinosus and gracilis grafts using standard MRI PACS for all included patients. Data of the actual graft sizes used intraoperatively were also extracted.

Results

We found that the graft diameter can be predicted in a numerically continuous manner as the square root of 2 * (AB + CD), where A and B refer to the semitendinosus cross-sectional length and breath respectively, and C and D refer to the gracilis cross-sectional length and breath respectively. The Pearson’s correlation coefficient between the predicted and actual graft diameter was 0.661 (p < .001), which shows a moderate positive correlation. Our method yields a high specificity of 92.6% and a moderate sensitivity of 67.2% if we define an adequate actual graft size as more than or equal to 8mm. A logistic regression model was performed is significant (p < .001), with the odds of the actual graft diameter being adequate increases by 12.8 for each additional mm of the predicted graft diameter (95% CI [5.2, 38.2]), and an area under receiver-operating characteristic (ROC) curve plotted with the respective logistic regression models shows good discrimination (AUC = 0.856).

Conclusions

We have presented a modified, practical method to predict the ACL hamstring graft size with high specificity using pre-operative MRI measurements that does not require any specialized software or methods, and can be reliably done even by junior members of the surgical team.