The present study aimed to evaluate the clinical efficacy of Andersson lesion (AL) treatments and prognostic factors using medium-to long-term follow-up data and discuss the clinical characteristics and treatment of AL. Forty-eight consecutive AL cases at our center from June 2011 to March 2018 were retrospectively analyzed, including 13 cases treated conservatively and 35 treated surgically. Epidemiological characteristics, treatment modalities, clinical features and outcomes, and prognostic factors of the Japanese Orthopaedic Association (JOA) recovery rate were reviewed. Neurological functional recovery was evaluated by American Spinal Injury Association (ASIA) classification. Clinical efficacy was evaluated by JOA score, visual analog scale (VAS) score, and Cobb's angle. The mean overall follow-up duration was 44.5±18.5 months (range, 27-85 months). There were 36 male and 12 female patients, with a mean age of 49.4±13.1 years (range, 26-72 years). The most common lesion location was the thoracolumbar region, i.e., T10-L2 (n=33; 68.8%), followed by the thoracic (n=10) and lumbar (n=5) regions. Patients treated surgically showed significantly better JOA scores, VAS scores and Cobb's angles at the final follow-up than did patients treated conservatively (P<.05). Univariate and binary logistic regression analyses identified two prognostic factors of the JOA score recovery rate: treatment modality (OR=0.157; 95%CI, 0.028-0.89; P=.036) and bone fusion (OR=9.965; 95%CI, 2.052-48.387; P=.004). Conservative treatment and bone nonunion predict worse JOA score recovery. Surgery remains the optimal treatment for AL in ankylosing spondylitis patients, with better clinical efficacy demonstrated by medium-to long-term follow-up data.