Instability of Depression Severity at Intake as a Moderator of Outcome in the Treatment for Major Depressive Disorder

Initial treatment for depression may not result in remission in up to two thirds of patients [1]; therefore, pre-treatment predictors of outcomes are important in determining the optimal treatment for each patient. Previous studies have explored patients' baseline characteristics as moderators that may identify for whom and under what circumstances distinct treatments have the most optimal effects. However, these studies have been largely unsuccessful in predicting treatment outcome.[2] The present study examined whether changes in the severity of depression during intake may predict treatment outcome. Study setting and participants have been previously described [3]. Patients diagnosed with MDD (n=156) were randomized to one of three treatment conditions: supportive-expressive therapy (SET, n=51), antidepressant medication (MED, n=55) or pill-placebo (PBO, n=50). Patients were excluded from randomization for not meeting MDD criteria (n=74), Hamilton Rating Scale for Depression (HRSD)<14 on first (n=4) or second evaluation, performed a week later (n=1), presence of primary disorders other than depression (n=45), and high suicide risk (n=9); (patients could meet multiple criteria). Depressive symptoms were assessed with the HRSD[4].The difference between the first and second pre-treatment evaluations served as an operationalization of pre-treatment changes in depressive symptoms. Treatments were provided for a total of 16 weeks. HRSD assessments throughout active treatment were conducted at weeks 2, 4, 6, 8, 12, 15 and 16 and served as continuous measures of outcomes. Multilevel modeling was performed using PROC MIXED in SAS. A logarithmic transformation (loge[week+1]) was used to quantify change. We conducted two longitudinal models to examine the effect of changes in pre-treatment HRSD scores (PT-HRSD) on outcomes. The first model tested the effect of PT-HRSD on changes in outcome over time for the entire cohort while the second examined differences between treatment conditions. In both models, we controlled for pre-treatment levels of depression severity. Mean age of the study cohort was 37.5 years (SD = 12.2); 92 patients were female. No baseline differences in demographic and clinical characteristics were found between treatment conditions. Mean pretreatment HRSD score was 20.3±4.06; seventeen patients showed identical HRSD scores in the two evaluations, while scores decreased in 87 patients and increased in 52 (mean change -0.9±3.5). When evaluating whether PT-HRSD instability predicted symptomatic change throughout treatment, we found that the omnibus test for the two-level interaction between time and PT-HRSD in predicting changes in outcomes was significant, B=-0.11, SE=.04, t(788)=-2.65, p=.008). Specifically, the estimated slope of the outcomes throughout the course of treatment was -2.69 when PT-HRSD=0 (i.e., stable pre-treatment depressive symptoms). A decrease in pre-treatment depressive symptoms (PT-HRSD) by one standard deviation (which is equal to 3.5 HRSD points), predicts a less rapid reduction in symptoms throughout treatment (an increase of B=0.41 in the negative slope of change). We next evaluated whether the ability of PT-HRSD to predict outcomes differs among the treatments. The omnibus test for the three-level interaction between time, PT-HRSD, and type of treatment in predicting outcomes was significant, F(2,785)=4.30, p=.013. Specifically, the effect of PT-HRSD on the slope of changes in outcomes over time for MED and SET was found to be significantly different from PBO, B=0.23, SE=.10, t(785)=2.18,p=.02 for MED, and B=0.33, SE=.11, t(785)=2.81, p=.005 for SET. Specifically, patients whose HRSD scores decreased by 1.15 points or more responded better to SET than PBO. Similarly, patients whose scores decreased by at least 0.7 points responded better to MED than PBO. As can be seen in Table 1, the largest effect of PT-HRSD on the slope of changes in outcomes over time was in the placebo condition. Table 1 Treatment response in patients who showed a decrease or increase in pre-treatment depression stability (PT-HRSD) and slope estimates for the within condition effect of PT-HRSD on changes in depressive symptoms throughout treatment Previous reports regarding predictors of treatment outcome for MDD found that early improvement throughout treatment was a predictor of stable response.[5] However, to the best of our knowledge, changes in severity of depression during the initial assessment of depressed patients have not been evaluated as predictors of treatment outcome. In the current study, involving patients meeting MDD diagnosis and having HRSD scores of 14 and above at two separate measurement points one week apart, pre-treatment changes in depressive symptoms significantly predicted symptomatic change throughout the course of treatment, even after controlling for pre-treatment severity of depression. Importantly, although the original analyses of these data demonstrated no overall differences in treatment outcome between the treatment groups [3], significant differences in the response to the two active treatments (psychotherapy and medication) versus control were apparent when using pre-treatment depression stability as a moderator. If validated in future studies, pre-treatment changes in depressive symptoms may be used to predict treatment outcomes and aid in treatment selection. Interestingly, the largest effect of pre-treatment depressive severity stability on outcomes was in the placebo condition. This suggests that instability of depression at intake is a highly sensitive predictor of the placebo versus medication and psychotherapy responses, and may help determine the magnitude of the placebo effect throughout treatment. Further studies are required in order to validate these findings and examine the underlying mechanism of this association.