Background. There have been no previous studies of the outcome of different neurotic disorders in which a prospective group with original randomization to treatment have been followed up over a long period. Such studies are important in identifying the factors associated with good and poor outcome.

Methods. A 5-year follow-up assessment was made of a cohort of 210 psychiatric out-patients seen in general practice psychiatric clinics with a DSM-III diagnosis of generalized anxiety disorder (71), panic disorder (74) or dysthymic disorder (65) and randomized to drug treatment, cognitive and behaviour therapy, and self-help. A total of 182 of the patients (87%) were assessed after 5 years by examination of hospital and GP records using a standardized procedure and outcome determined with a four-point outcome scale.

Results. One hundred and seven (60%) of the patients had a favourable outcome but the remainder continued to be handicapped either intermittently or continuously throughout the 5-year period. Analysis of the value of initial data in predicting outcome using polychotomous step-wise logistic regression revealed that five variables were significant predictors of poor prognosis: older age; recurrent episodes; the presence of personality disorder at entry; general neurotic syndrome at entry; and symptom severity after 10 weeks. The initial DSM diagnosis and original treatment given, together with ten other variables, were of no predictive value.

Conclusions. The long-term outcome of neurotic disorder is better predicted by age, personality and recency of onset than by other clinical variables with the exception of initial response to treatment.