We examined predictors of outcome 1 year after completion of a randomi
zed clinical trial assessing the additive efficacy of two forms of exp
osure with response prevention to a core of cognitive-behavioral thera
py (CBT) for bulimia nervosa (BN). One hundred one women who met DSM-I
II-R criteria for BN, and who completed the clinical trial, were avail
able for follow-up at 1 year. Predictor variables were assessed prospe
ctively and partitioned temporally to reflect lifetime history (includ
ing personality), pretreatment clinical status, and posttreatment clin
ical status. Outcome was based on the frequency of binging and purging
in the 3 months before assessment based on carefully constructed life
chart interviews. A series of stepwise logistic regressions were perfo
rmed to determine independent predictors of 1-year outcome while contr
olling for treatment received. Demographic variables were unrelated to
treatment outcome. A history of obesity was predictive of poor outcom
e, whereas a history of alcohol dependence decreased the odds of poor
outcome, High self-directedness on the Temperament and Character Inven
tory (TCI) predicted favorable outcome at 1 year, whereas personality
disorder symptoms were not predictive. Pretreatment global functioning
, bulimia scores on the Eating Disorders Inventory (EDI), and the pres
ence of major depression predicted poor outcome. Posttreatment binging
, food restriction, and urges to binge on a cue reactivity assessment
predicted poor outcome at 1 year. The character trait of self-directed
ness is a strong predictor of good outcome for CBT, and methods to enh
ance this trait may be worthy of investigation. Low global functioning
and the presence of major depression at presentation may require addi
tional treatment than focused CBT for BN. Our results argue for treatm
ent goals that include abstinence from binging and restricting and dec
reases in urges to binge in response to high-risk cues. Copyright (C)
1998 by W.B. Saunders Company.