Patients with cervical dystonia who are non-responders to Botulinum toxin qualify for surgery. Selective peripheral denervation (Bertrand's procedure, SPD) and deep brain stimulation of the globus pallidus (GPi-DBS) are available surgical options. Although peripheral denervation has potential advantages over DBS, the latter is nowadays more commonly performed. We describe the long-term outcome of selective peripheral denervation as compared with GPi-DBS, along with the findings of literature review. Twenty patients with selective peripheral denervation and 15 with GPi-DBS were included. Tsui scale, a visual analogue scale, and the global outcome score of the Toronto Western Spasmodic Torticollis Rating Scale were used to define a "combined global surgical outcome". The "combined global surgical outcome" for patients with selective peripheral denervation or pallidal stimulation was respectively "bad" for 65 and 13.3 %, "fair-to-good" for 30 and 26.7 %, and "marked" improvement for 5 and 60 % (p < 0.001). Improvement on visual analogue scale (p < 0.002), global outcome score (p < 0.002), and Tsui score (p < 0.000) was larger for the pallidal stimulation group. Seventy-five percent of patients with selective peripheral denervation and 60 % of patients with pallidal stimulation reported side effects. Seven patients with selective peripheral denervation successively underwent GPi-DBS, with a further significant improvement in the Tsui score (-48.6 ± 17.4 %). GPi-DBS is to be preferred to selective peripheral denervation for the treatment of cervical dystonia because it produces larger benefit, even if it can have more potentially severe complications. GPi-DBS is also a valid alternative in case of failure of SPD.