Background: The definition of outcome in antidepressant treatment stud
ies may be a crucial factor influencing the relationship between possi
ble treatment variables and treatment response. We therefore wanted to
investigate possible relationships between baseline severity of depre
ssion and anxiety, and different definitions of outcome among outpatie
nts with major depressive disorder undergoing antidepressant treatment
. Methods: Two hundred and forty-eight patients diagnosed with major d
epression with the Structured Clinical Interview for DSM-III-R Diagnos
is - Patient Edition were treated with fluoxetine 20 mg/day for 8 week
s. Patients were evaluated both pretreatment and posttreatment with th
e 17-item Hamilton Rating Scale for Depression (Ham-D-17), the Clinica
l Global Impressions Scales for Severity (CGI-S) and Improvement (CGI-
I). We chose four continuous definitions of response using a linear re
gression method to analyze the relationship to baseline anxiety and de
pression. We used a logistic regression analysis for the relationship
between seven categorical definitions of response and baseline severit
y of depression and anxiety. Results: Greater endpoint severity was si
gnificantly positively related to greater baseline severity of depress
ion and anxiety. Lesser baseline severity of depression or anxiety pre
dicted a greater degree of response on either the endpoint CGI-I score
(with lower scores indicating greater improvement) or the percent cha
nge in HAM-D-17 score (with higher percent change indicating greater i
mprovement). For all seven categorical definitions of response, lower
baseline scores were significantly related to the probability of being
a responder. Conclusions: These findings support the impression that
how outcome is defined affects the strength and direction of observed
relationships with predictive variables. Methodological implications a
re discussed.