Abstract It is generally believed that opioids relieve neuropathic pain less effectively than nociceptive pain and that they have no effect on some of the key characteristics of neuropathic pain such as touch-evoked pain (allodynia). Tramadol is an analgesic drug acting directly on opioid receptors and indirectly on monoaminergic receptor systems. The aim of this trial was to determine whether tramadol relieved painful polyneuropathy and reduced allodynia. The study design was randomised, double-blind, placebo-controlled and cross-over. After baseline observations, 45 patients were assigned to one of the two treatment sequences. The dose of tramadol slow-release tablets was titrated to at least 200 mg/day and at highest 400 mg/day. During the two treatment periods of 4 weeks duration, patients rated pain, paraesthesia and touch-evoked pain by use of 0–10 point numeric rating scales. Mechanical allodynia induced by stimulation with an electronic toothbrush was rated at the end of each treatment period with a similar scale. Thirty-four patients completed the study. Their ratings for pain (median 4 vs. 6, P=0.001), paraesthesia (4 vs. 6, P=0.001) and touch-evoked pain (3 vs. 5, P<0.001) were lower on tramadol than on placebo, as were their ratings of allodynia (0 vs. 4, P=0.012). The number needed to treat to obtain one patient with ≥50% pain relief was 4.3 (95% confidence interval 2.4–20). It is concluded that tramadol appears to relieve both ongoing pain symptoms and the key neuropathic pain feature allodynia in polyneuropathy.