Guillain-Barré syndrome (GBS) is an acute inflammatory disorder of the peripheral nervous system thought to be due to autoimmunity for which immunotherapy is usually prescribed. To provide the best evidence on which to base clinical practice, we systematically reviewed the results of randomized trials of immunotherapy for GBS. We searched the Cochrane Library, MEDLINE and EMBASE in July 2006 and used the methods of the Cochrane Neuromuscular Disease Group to extract and synthesize data. Almost all trials used a 7-point disability grade scale. In four trials with altogether 585 severely affected adult participants, those treated with plasma exchange (PE) improved significantly more on this scale 4 weeks after randomization than those who did not, weighted mean difference (WMD) -0.89 (95% confidence interval (CI) -1.14 to -0.63). In five trials with altogether 582 participants, the improvement on the disability grade scale with intravenous immunoglobulin (IVIg) was very similar to that with PE, WMD -0.02 (95% CI -0.25 to 0.20). There was also no significant difference between IVIg and PE for any of the other outcome measures. In one trial with 148 participants, following PE with IVIg did not produce significant extra benefit. Limited evidence from three open trials in children suggested that IVIg hastens recovery compared with supportive care alone. Corticosteroids were compared with placebo or supportive treatment in six trials with altogether 587 participants. There was significant heterogeneity in the analysis of these trials which could be accounted for by analysing separately four small trials of oral corticosteroids with altogether 120 participants, in which there was significantly less improvement after 4 weeks with corticosteroids than without, WMD -0.82 (95% CI -0.17 to -1.47), and two large trials of intravenous methylprednisolone with altogether 467 participants, in which there was no significant difference between corticosteroids and placebo WMD -0.17 (95% CI 0.06 to -0.39). None of the treatments significantly reduced mortality. Since approximately 20% of patients die or have persistent disability despite immunotherapy, more research is needed to identify better treatment regimens and new therapeutic strategies.