A previous study showed that most ankylosing spondylitis (AS) patients presented recurrence within 6 months post-discontinuation of etanercept. How to reduce recurrence following discontinuation of etanercept should be further researched. In this study, 111 ankylosing spondylitis patients meeting the Assessment in AS 20 % response (ASAS20) criteria after 12-week administration of etanercept were randomized into three groups: Group I, 150 mg thalidomide once/day; Group II, 1 g sulfasalazine, twice/day; Group III, NSAIDs for the maintenance treatment. The patients were regularly followed up once a month, and AS recurrence was evaluated with the Bath Ankylosing Spondylitis Disease Activity Index (BASDAI), the Bath Ankylosing Spondylitis Functional Index (BASFI), the patient global assessment (PGA), and rachialgia. The follow-up lasted for 1 year, and AS recurrence was considered as the end of a visit. Finally, 100 patients completed the follow-up study, of whom 30 were in Group I, 33 in Group II, and 37 in Group III. The average follow-up period was 5.1 ± 3.9 months and the longest lasted for 12 months. At the end of the follow-up study, the recurrence rates in Groups I, II, and III were, respectively, 60.0 % (18/30), 84.8 % (28/33), and 89.2 % (33/37). The recurrence rates of Group I were statistically significantly lower than that of Group II and III (P = 0.0265; P = 0.0053), while there was no significant difference between Group II and Group III. In addition, we found that PGA, C-reactive protein (CRP), and spinal inflammation could be regarded as predictive factors for AS recurrence by analysis with the Cox proportional hazard model. This study points to a new way for maintenance therapy of AS following discontinuation of etanercept and reveals several useful indicators for prediction of AS recurrence.