Identifying Acute Lumbar Spondylolysis in Young Athletes with Low Back Pain Retrospective Classification and Regression Tree Analysis

Study Design. Case-control study. Objective. The aim of this study was to establish an algorithm to distinguish acute lumbar spondylolysis (LS) from nonspecific low back pain (NSLBP) among patients in junior high school by classification and regression tree (CART) analysis. Summary of Background Data. Rapid identification of acute LS is important because delayed diagnosis may result in pseudarthrosis in the pars interarticularis. To diagnose acute LS, magnetic resonance imaging (MRI) or computed tomography is necessary. However, not all adolescent patients with low back pain (LBP) can access these technologies. Therefore, a clinical algorithm that can detect acute LS is needed. Methods. The medical records of 223 junior high school-aged patients with diagnosed acute NSLBP or LS verified by MRI were reviewed. A total of 200 patients were examined for establishing the algorithm and 23 were employed for testing the performance of the algorithm. CART analysis was applied to establish the algorithm using the following data; age, sex, school grades, days after symptom onset, history of LBP, days of past LBP, height, passive straight leg raising test results, hours per week spent in sports activities, existence of spina bifida, lumbar lordosis angle, and lumbosacral joint angle. Sensitivity and specificity of the algorithm and the area under the ROC curve were calculated to assess algorithm performance. Results. The algorithm revealed that sex, days after symptom onset, days of past LBP, hours per week spent in sports activities, and existence of spina bifida were key predictors for identifying acute LS versus NSLBP. Algorithm sensitivity was 0.64, specificity was 0.92, and the area under the ROC curve was 0.79. Conclusion. The algorithm can be used in clinical practice to distinguish acute LS from NSLBP in junior high school athletes, although referral to MRI may be necessary for definitive diagnosis considering the algorithm's sensitivity.