Risk Factors for Sustaining a Second ACL Injury after Primary ACL Reconstruction in Female Football Players: A Study Investigating the Effects of Follow-Up Time and the Statistical Approach

Abstract

Abstract Background Studies evaluating risk factors for sustaining an anterior cruciate ligament (ACL) injury have different, sometimes contrasting, results. Different follow-up times and statistical approaches may be a reason for these differences. The aim of this study was to explore if different follow-up times and statistical approaches, classification and regression tree (CART) analysis and Cox regression, would impact on the association between various candidate risk factors and ACL injury in female football players. In total, 112 active female football players, 18 ± 8 months after ACL reconstruction (mean age ± SD, 20 ± 2 years), were included and followed for at least 36 months. At baseline, all players underwent assessment of range of motion of knee and ankle joints, functional tests, and answered questionnaires regarding knee function, psychological and personality traits. Nineteen independent variables were included for the CART analysis and for univariable Cox regression and compared using four different follow-up times: 0–12, 0–24, 0–36, and 0–>36 months. Results Forty-three (38%) players sustained a second ACL injury. The identified risk factors varied depending on follow-up time both with CART analysis and with Cox regression. CART identified 12 of the 19 independent variables and selected between 5 and 6 of the variables in the four different follow-up times associated with second ACL injury. The accuracy of the different follow-up times for the CART varied between 86 and 93% with 77–96% sensitivity and 70–81% specificity. Cox regression identified two risk factors: knee extension at 0–36 months and 0–>36 months, and time between primary injury and surgery at 0–>36 months. The accuracy varied between 54 and 64% with 44–88% sensitivity and 32–71% specificity. Conclusions The identified risk factors associated with a second ACL injury varied depending on the follow-up time and statistical approach used. Thus, in future research on risk factors, the time athletes are followed up and the type of statistical methods used are important to discuss.

Keywords

• Cox regression
• CART
• Football
• Prediction
• Screening