BACKGROUND: The purpose of the study was to evaluate time trends and effects of co-existing autism spectrum disorders (ASD) on outcome in an ongoing long-term follow-up study of anorexia nervosa (AN). METHODS: The Morgan-Russell Outcome Assessment Schedule (MROAS) was used at 6-, 10- and 18-year follow-up of a representative sample of 51 individuals with teenage-onset AN and a matched group of 51 healthy comparison cases. The full multinomial distribution of responses for the full scale and each of the subscales was evaluated using exact nonparametric statistical methods. The impact of diagnostic stability of ASD on outcome in AN was evaluated in a dose-response model. RESULTS: There were no deaths in either group. Food intake and menstrual pattern were initially poor in the AN group but normalised over time. MROAS 'mental state' was much poorer in the AN group and did not improve over time. The psychosexual MROAS domains 'attitudes' and 'aims' showed persistent problems in the AN group. In the MROAS socioeconomic domain, the subscales 'personal contacts', 'social activities' and 'employment record' all showed highly significant between-group differences at all three follow-ups. A statistically significant negative dose-response relationship was found between a stable diagnosis of ASD over time and the results on the subscales 'mental state', 'psychosexual state' and 'socio-economic state'. CONCLUSIONS: Outcome of teenage-onset AN is favourable with respect to mortality and persisting eating disorder, but serious problems remain in the domains 'mental state', 'psychosexual function' and 'socioeconomic state'. Outcome is considerably worse if ASD is present. Treatment programmes for AN need to be modified so as to accommodate co-existing ASD.