We are grateful to Drs Wood & Flower for contributing to the debate on treatment setting in adolescent anorexia nervosa. Our aim was indeed to open rather than close discussion.

We agree that it is of paramount importance that anorexia nervosa is seen for the serious condition with high morbidity and mortality that we know it to be, rather than the trivial disorder sometimes portrayed by the media. It is right, however, for commissioners to expect an evidence-based case for expensive treatments. Despite the questions raised in our paper, our service treats a large number of adolescents as in-patients and continues to make and support significant numbers of referrals to specialist eating disorder in-patient services.

Nevertheless, it is extraordinary that the following questions are so rarely addressed:

1. (a) Could it be that in-patient treatment has negative (side-)effects?

2. (b) Could there be some intrinsic features of anorexia nervosa, such as ineffectiveness, low self-esteem or past history of abuse, which might make those with anorexia nervosa particularly vulnerable to these negative effects?

3. (c) Might these negative effects sometimes outweigh the benefits?

We would take these questions for granted in evaluating a new drug therapy.

The point Drs Wood & Flower make about systemic factors as predictors of outcome is an important one that our group has previously researched (Reference Gowers and NorthGowers & North, 1999). Where there is family or social difficulty, however, does this mean that the adolescent is better treated within or outwith the family home? Does this difficulty add to the case for admission or the case against? In view of the high rates of relapse after weight restoration in hospital, we contest that one could from testable hypotheses either way.

The National Health Service Executive has rightly judged that further evidence of the effectiveness of treatment in different settings is required. We are pleased to report that our group was awarded a Health Technology Assessment grant to conduct a randomised controlled trial of treatment setting covering the north-west of England. We hope in the course of the 4-year pragmatic study to contribute to the debate on when specialist eating disorder in-patient units may be helpful and for whom. We are also examining family satisfaction and acceptability. Of course, this large study will not provide the last word on the issue, but we must avoid the negativism which suggests it is better not to carry out research in case the results are misinterpreted.

Almost certainly in-patient admission sometimes saves lives. Nevertheless, almost all series show high rates of relapse after discharge (Reference Crisp, Norton and GowersCrisp et al, 1991; Reference Eisler, Dare and RussellEisler et al, 1997) and however loaded with poor prognostic features our series may have been, a good outcome for 3 out of 21 is very poor.