To the Editor: As clinicians and researchers, we found the article by D. Blake Woodside, M.D., F.R.C.P.C., et al. (1) of considerable interest because it is a naturalistic study focused on a problematic issue. Indeed, studying predictive factors of the failure of therapeutics for the most severely ill anorexia nervosa patients has been rare (2–4).

We identified three criteria that influence the dropout rate from inpatient programs. The first criterion has previously been described: the treatment method (2, 5). The second criterion is the definition of "dropout" itself. Indeed, none of these studies (1–4) considered "dropout" identically. As a result, the dropout rates varied from 20% if one considers only patient-initiated discharges (4) to 50% if one considers patient-initiated discharges and staff-initiated discharges (1). The third criterion is the clinical characteristics of the subjects studied. The previous studies (1–4) focused on adults with anorexia nervosa ages 20.5 years (SD=4.8) to 27.1 years (SD=9.0), on average. To our knowledge, there has been no study about this issue in adolescents.

In our adolescent inpatient unit in France, 268 girls (mean age=16.7 years, SD=2.0) were hospitalized between 1996 and 2004. The mean duration of illness was 20.4 months (SD=17.1, median=13.9). When we considered dropouts as all subjects who did not achieve their therapeutic contract (did not reach the target weight for discharge) (6), we found a dropout rate of 24.6% (N=66). One-half were staff-initiated, and one-half were patient-initiated. Only six of 66 (9%) were discharged at the first part of their therapeutic contract (early dropouts). As usual, with the youngest patients, the purging subtype was rare (N=51, 19%). To examine dropout predictors, we performed a step-by-step backward logistic regression analysis. We considered 13 predictive variables reported by the literature (1–4): body mass index (kg/m²) at admission, maximum and minimum previous body mass indexes, age at admission, length of hospitalization, time since first treatment, duration of illness, anorexia nervosa subtype, number of previous hospitalizations, body mass index at discharge, age at onset, educational status, and socioeconomic status.

Four variables were significantly related to dropout: higher body mass index at admission (odds ratio=1.5, p<0.03), lower body mass index at discharge (odds ratio=0.2, p=0.0001), longer length of hospitalization (odds ratio=1.0, p=0.0001), and later age at onset (odds ratio=1.4, p=0.005).

In contrast to the report by Dr. Woodside and colleagues, our dropouts had a longer length of hospitalization. This could be explained by the setting of our therapeutic contract. Indeed, the staff never discharge a patient because of lack of progress (e.g., lack of weight gain) in the first weeks.

The few patient-initiated discharges (N=27, 10%) were probably due to the subjects’ age: they can leave the hospital only with their parents’ permission. One may hypothesize that the therapeutic alliance between the parents and staff helps the subjects remain hospitalized.

Although most of the studies (1, 3, 4) indicated that the purging subtype of anorexia nervosa was a predictor of dropout, we did not find this result (approximately 19% had the purging subtype in both groups). Once more, the setting of our inpatient program and the adolescents’ age could explain this result. These results stress the need for further research on this issue.