S ingle follow-up studies have identified a low body weight at referral as a poor prognostic indicator for intermediate or long-term follow-up (1, 2). Morgan and Russell (3) observed an association between low body weight (less than 60% average body weight, roughly equivalent to a body mass index of 12.5 kg/m 2 in females aged 17-19 years) at referral and poor general outcome after 5 years. In a previous study (4), we found that 24 patients with body mass indexes less than 13 kg/m 2 at referral had significantly lower body mass indexes at follow-up than 57 patients who had higher body mass indexes at referral; the mean follow-up period was 11.7 years. Possible explanations for the discrepancies regarding the prognostic implications of a low body weight at referral include small sample sizes and crude categorizations of the weight outcome.The hypothesis of the present study was that a low body mass index at referral predicts a low body mass index at long-term follow-up. To test this hypothesis we assessed body mass indexes at referral and at follow-up in a large study group (N=272) to determine their interrelationship and to assess the influence of age at onset, age at referral, and age at follow-up, respectively.
METHODThe study group was a composite of five study cohorts with a total of 341 consecutively ascertained inpatients with anorexia nervosa. Twenty-four of the patients were males. The ascertainment, diagnostic assessment according to DSM-III-R, and outcome of these study cohorts have been described in detail elsewhere (4-9). For the present analysis, all 24 of the males, seven females with additional somatic diseases at referral, and seven pretreated females whose body mass indexes at referral were greater than 17.5 kg/m 2 were excluded; 31 additional females could not be traced at follow-up. The final study group included 272 females whose body mass indexes at referral and body mass indexes at follow-up or time of death were available. Their mean ages at referral and at follow-up were 16.7 years (SD=4.5, range=10-42) and 26.2 years (SD=6.9, range=15-58), respectively. Mean length of follow-up was 9.5 years (SD=5.3, range=0-33.6) and included the survival times of the 12 patients who had died after a mean of 4.2 years (SD=4.0, range=0-13). Deaths were related to the sequelae of emaciation in 10 patients and to suicide in two patients.The patients' body mass indexes at referral were plotted against their body mass indexes at follow-up and Pearson's correlation was calculated. Body mass indexes were plotted in 1-kg/mg 2 intervals from 9 to 27 kg/mg 2 . Patients with low body mass indexes at referral were compared with patients with higher body mass indexes at referral by using repeated U tests. Significance was adjusted for multiple testing (10). We also charted the proportion of patients whose body mass indexes at follow-up were 17.5 kg/m 2 or less, those whose body mass indexes were in the fifth percentile or below, and those whose body mass indexes were in the 10th percentile or below. Body mass index ...