Fifty-five subjects (31%) were classified as full responders; 75 (42.9%) as partial; and 45 (25.7%) non-responders. Of the full responders, half (n =28, or 16% of the study group) were remitters (HRSD ≤6). Race was the only demographic characteristic that differed significantly (χ2 =10.99, p =0.004) between partial and non-responders, with 85% (n =47) of full responders, 93% (n =70) of partial responders, and 71% (n =32) of non-responders self-identified as white (). The remaining 29% (n =13) of non-responders all self-identified as African-American. Socioeconomic status, as estimated by the Hollingshead Index of Socioeconomic Status, was evaluated in 163 of the 175 participants, and no significant difference between the three groups was detected. African-American participants did not differ in terms of severity of depression or anxiety, social support, or past history of substance use when compared to white participants. However, African-Americans reported missing significantly more doses of escitalopram as compared to white participants: 35% vs 8% were non-adherent (χ2 =14.79, p =0.0001), with adherence defined as having at least 80% of visits with no missed dosages reported.
Baseline demographic characteristics of study participants
Clinically, severity of depression and anxiety symptoms, hopelessness, health-related quality of life (disability), and social support differed across the three groups (). Partial responders reported higher levels of hopelessness than full responders, full responders reported more social support and better health-related quality of life than non-responders. Additionally, both full and partial responders had less severe depressive and anxiety symptoms at baseline compared to non-responders. Co-existing medical burden (CIRS-G), duration of index depressive episode, age of lifetime onset, percent with recurrent depression, suicidal ideation, and cognitive function (including executive impairment as measured by the Initiation/Perseveration subscale of the Mattis Dementia Rating Scale) did not differ across the three groups. Rates of pre-treatment benzodiazepine usage were 44% in nonresponders, 35% in partial responders, and 20% in full responders (
Baseline clinical and neuropsychological characteristics of study participants
Self-esteem was most impaired in non-responders and was the only contributor to the differences in social support (ISEL) between full and non-responders (F =4.51, df =2,169 p =0.012). Self-appraised social support, belonging, and tangible support did not differ. Full and partial responders had lower psychological and somatic anxiety subscores of the Hamilton Anxiety Scale (F =6.36, df =2,172 p =0.002 for psychological subscore; F =5.77, df =2,172 p =0.004 for somatic subscore). However, the groups did not differ in rates of any coexisting anxiety disorder (on the SCID for DSM-IV), including generalized anxiety disorder, or substance use.
In order to determine which variables predicted treatment response, we performed a hierarchical polytomous logistic regression. There was no effect of demographic variables (including race) on response in the first step (Likelihood-rated (LR) test
=0.50). The addition of clinical variables was highly significant (LR
<0.005): self esteem (Wald χ2
=0.024, Odds Ratio [OR] =0.89 [0.79, 0.98]) and severity of anxiety symptoms (χ2
=0.023, OR =1.09 [1.01,1.18]) were retained as reliable predictors in the model. Predicted versus observed response was concordant in 65% of cases, discordant in 34% and tied in 1%. Race and self-reported medication adherence were not significant predictors in the multivariate analysis.