We reviewed 200 patients with paraneoplastic encephalomyelitis (PEM) and anti-Hu antibodies to show possible clinical differences with respect to previous series, and to identify patient, tumour and treatment-related characteristics associated with neurological disability and survival. The median age of the 200 patients was 63 years (range 28-82 years) and 75% were men. The predominant neurological syndromes were sensory neuropathy (54%), cerebellar ataxia (10%), limbic encephalitis (9%) and multifocal involvement (11%). Sensorimotor neuropathies with predominant motor involvement were observed in only 4% of the patients. Pathological or X-ray evidence of a tumour was obtained in 167 patients (83%) and was a small-cell lung cancer (SCLC) in 74% of those with histological diagnosis. Coexistence of extrathoracic tumours with SCLC was rare (0.5%). Positive Hu immunoreactivity was observed in the extrathoracic tumours of six out of seven patients in whom autopsy or long-term follow-up ruled out a coexisting SCLC. PEM preceded the diagnosis of the tumour in 71% of patients (mean delay +/- SD 6.5 +/- 7.0 months; range 0.1-47 months). In the 24 patients in whom the tumour diagnosis was the initial event, PEM predicted the progression or relapse of the tumour in 87% of them. No tumour was found in 33 patients, including four who had a post-mortem study and four with >5 years of follow-up. In a logistic regression analysis, treatment of the tumour, associated or not with immunotherapy, was an independent predictor of improvement/stabilization of PEM [odds ratio 4.56; 95% confidence interval (CI) 1.62-12.86]. Cox multivariate analysis indicated that the variables independently associated with mortality were: age >60 years [relative risk (RR) 1.49; 95% CI 1.05-2.12], Rankin score at diagnosis >3 (RR 1.60; 95% CI 1.12-2.28), more than one area of the nervous system affected (RR 1.61; 95% CI 1.08-2.40), and absence of treatment (RR 2.56; 95% CI 1.76-3.71). We conclude that, unlike previous series, the majority of our patients were male, and there was a low occurrence of predominantly motor neuropathies and extrathoracic tumours coexisting with SCLC. When the diagnosed extrathoracic tumour expresses Hu antigens, further tests to rule out a coexisting SCLC are probably unnecessary. Finally, the predictors of mortality and PEM evolution found in the study may be important in the design of future therapeutic protocols, and emphasize the importance of early diagnosis and treatment of the underlying tumour.