BACKGROUND The optimal treatment of Graves disease (GD) is still controversial. Surgery is one treatment option along with radioactive iodine (RAI) and antithyroid medication. In this evidence-based review, we examine four issues: (1) Is surgery better than RAI or long-term antithyroid medication? (2) What is the recommended surgical approach? (3) How does the presence of Graves' ophthalmopathy (GO) influence the role of surgery? (4) What is the role of surgery in children with GD? METHODS We conducted a systematic review of the literature using evidence-based criteria regarding these four issues. RESULTS (1) There are no recommendations reaching any grade of evidence for which treatment to choose for adults with GD. (2) Total thyroidectomy has complication rates equal to those seen with lesser resections but it has higher cure rates and negligible recurrence rates (Level I-IV data leading to a grade A recommendation). (3) Data support surgery when severe GO is present, but RAI combined with glucocorticoids may be equally safe (Level II-IV data, grade B recommendation). The extent of thyroid resection does not influence the outcome of GO (Level II data, grade B recommendation). (4) Based on the available data, definitive treatment can be advocated for children (Level IV data, grade C recommendation) using either RAI or surgery. No recommendation can be given as to whether RAI or surgery is preferred owing to the lack of studies addressing this issue. Increased cancer risk with RAI in children below the age of 5 years supports surgery in this setting (Level I data, grade A recommendation). CONCLUSIONS If surgery is considered for definitive management, evidence-based criteria support total thyroidectomy as the surgical technique of choice for GD. Available evidence also supports surgery in the presence of severe endocrine GO. Children with GD should be treated with an ablative strategy. Whether this is achieved by total thyroidectomy or RAI may still be debatable. Data on long-term cancer risk are missing or conflicting; and until RAI has proven harmless in children, we continue to recommend surgery in this group.