The authors analyse the results up to death in 103 followed-up patients undergoing unilateral percutaneous cervical cordotomy for persistent cervico-thoracic malignant pain (45 cases of Pancoast syndrome and 58 cases of thoracic pain associated with lung cancer or metastases). On the basis of epidemiological data, relationships emerge between onset of pain, stage of cancer, patient survival and lasting efficacy of pain relief. Twenty (44%) of 45 patients with Pancoast syndrome were pain-free up to death as a result of cordotomy alone, while only 13/58 patients (22%) with thoracic pain were pain-free as a result of cordotomy alone owing to the very high incidence of mirror pain in this group of patients (42/58 patients, 72%) compared to those with Pancoast syndrome (14/45 patients, 31%). The type and intensity of mirror pain, however, were of such a nature in both groups as to be amenable to control with analgesic drugs. In both groups of patients, there was a low incidence of the causes of post-cordotomy pain recurrence contralateral to the lesion, i.e., deafferentation pain, fading of analgesia, and pain above the levels up to which deep pin-prick analgesia had been obtained. Cordotomy alone or, as necessary, in conjunction with analgesic drugs afforded complete pain control in 34/45 patients (75%) with Pancoast syndrome and in 50/58 patients (86%) with thoracic pain. These data provide evidence of the unique usefulness of the procedure in controlling otherwise intractable persistent cervicothoracic malignant pain, when the technique is correctly performed.