Forty-five articles were included.
Methodological flaws in the primary studies included small sample size, RLS history unclear, concomitant therapy, errors in ascertainment of exposure to intervention, inclusion of patients with narcolepsy, non-systematic group assignment in studies described as randomised, short wash-out periods in cross-over trials, no statistical analyses, and vague end points and clinical criteria.
1. Dopaminergic medications:
Levodopa (8 double-blind trials ranging from 1 to 4 weeks duration, 107 patients): inconsistent results with some studies reporting no significant difference between treatments and others favouring levadopa over placebo for some outcome measures. 7 clinical series reported that the emergence of rebound or augmentation was the greatest problem.
Two double blind RCTs with 18 patients compared oxycodone (reported benefit from oxycodone compared to placebo) and proxyphene (no difference from placebo).
Benzodiazepines: two double blind cross-over studies, 1 to 4 weeks duration, 12 patients studied RLS and reported inconsistent results.
3. Anticonvulsant medications: Carbamazepine: one double blind RCT of five weeks duration and 181 patients reported carbamazepine was significantly better than placebo. A large placebo effect was noted. One RCT cross-over with six patients had no statistical analyses.
Gabapentine: two open label series with 22 patients neither of which reported statistical analyses.
4. Medications drawn from other classes: Clonidine: two double blind RCT, duration from 3 days to 3 weeks, 20 patients both reported significant benefit from clonidine compared to placebo.
Iron: one case-control with 18 patients reported benefit only for patients with ferritin <= 45 mg/l. Two case series with 45 patients treated with iron in the form of intravenous iron, oral iron and blood transfusion reported benefit from iron.
Non pharmacological therapy: Cognitive therapy vs clonazepam: one study, 4 weeks duration, 16 patients. Statistics were not clearly presented.