I fully agree with the authors that a careful definition of context is all important. So it was disappointing that they took my comment on being reassured out of the context in which it was placed; every death of a teenager is a tragedy. To note that the observed annual death rate for anorexia nervosa is around a tenth of that for the paediatric conditions they cite, and far less than the rate of death by suicide in young patients with schizophrenia, is not to be relaxed about risk in anorexia nervosa, nor to imply that specialised treatment for anorexia nervosa is not required.

In my response to Robinson's article I did not sufficiently well position my own observation about medical complications, which would have been better phrased as ‘the only common complication of clinical significance’. Morbidity, which I was seeking to distinguish from the biological response to starvation, sometimes goes beyond bone mineralisation problems, and often does when chronicity becomes established. However, unlike Wentz et al's prospective outcome study, ^{Reference Wentz, Gillberg, Anckarsäter, Gillberd and Råstam1} which reported no deaths over an 18-year follow-up period, the study of brain structure and function cited ^{Reference Chui, Christensen, Zipursky, Richards, Hanratty and Kabani2} was not based on a community sample. Their findings were derived from a cohort of acutely ill hospitalised adolescents, and were most evident in those who remained at low weight in their 20s.

I am grateful to learn about the Junior MARSIPAN report, and hope that most readers recognised that I was situating my response to Robinson's article from a community perspective, including how ‘risk’ may be constructed by those responding to newly presenting patients with anorexia.