Personalized Indications for CBT and Antidepressants in Treating Depression

Depression is among the most common and burdensome diseases worldwide and in Canada. It is treatable with effective pharmacological and psychological treatments, but fewer than 50% of individuals achieve remission with the first treatment. The unsatisfactory outcomes are at least partly due to a gap in evidence on which treatment benefits which individual.

The two most commonly used treatment modalities for depression are antidepressant medication and psychotherapy. However, only a minority of individuals with depression achieve remission with each type of treatment. On average, combination of antidepressants and psychotherapy achieves better results than either alone, but using two treatments at the same time may be wasteful and unnecessary. Several lines of evidence suggest that the apparently superior results of combination may be due to some individuals preferentially responding to antidepressants and others to psychological treatment. These findings point to the possibility that it may be possible to improve outcomes of depression by matching psychological and pharmacological treatment to individuals who are most likely to benefit from each one. The investigators propose to close this gap using a combination of clinical predictors and biomarkers to optimize the choice between psychological and pharmacological treatment.

The investigators have identified that a symptom profile with prominent reduction in interest and activity, history of maltreatment in childhood, increase in systemic inflammation, and a genetic variant increasing sensitivity to environment predict poor response to antidepressant medication. Indirect evidence suggests that the same factors may predict better response to cognitive-behavioural psychotherapy.

The investigators focus on these four predictors with substantial prior evidence:

Reduced interest and activity. A symptom dimension of interest and activity emerged as the strongest predictor of poor outcome of antidepressant medication treatment. This prediction replicated with undiminished effect size in the largest study of antidepressant treatment carried out to date,4 and additional independent replications confirmed its robustness. Cognitive behavioural therapy improves activity and interest better than antidepressant medication.

Childhood maltreatment. A meta-analysis established that a history of maltreatment in childhood predicts non-response to antidepressants but not to psychotherapy in adults with depression. History of maltreatment may actually predict greater benefits from psychological treatment.

Systemic inflammation. Inflammation may be a pathogenic mechanism in a proportion of depression cases. The invesigators found that increased C-reactive protein (CRP), a marker of systemic inflammation, predicted non-response to a commonly used antidepressant. Cognitive behavioural therapy can decrease CRP and may be the preferred treatment option for individuals with high levels of systemic inflammation.

Glucose metabolism. Insulin resistance, indicated by higher levels of insuline needed to maintain a normal blood glucose levels, is a marker of metabolic health risk and has been associated with response to psychotropic medications in the related condition bipolar disorder.

Combination of predictors. The above predictors are at least partly independent and may combine to a clinically significant prediction of treatment outcomes. Therefore, the investigators propose to evaluate the predictive validity of each predictor alone as well as all four predictors in combination.

Other predictors. The selection of the four predictors described above represents a balance between comprehensiveness and complexity, but it is not exhaustive. To maximize the potential for integrative analysis, the investigators will also evaluate addition potential predictors including anxiety, dysfunctional attitudes, and personality disorders.

The investigators propose to test whether these four predictors alone or in combination differentially predict the outcomes cognitive-behavioural psychotherapy and antidepressant medication with clinically significant accuracy. If this hypothesis is supported, the resulting predictor will allow personalized selection of treatment for depression, leading to improved outcomes and healthcare efficiency. Additional objectives include replication of additional predictors and integrative analyses aimed at refining the treatment choice algorithms.

Depression has been reclassified into two categories: major depressive disorder (MDD), and persistent depressive disorder (PDD), which may be associated with distinct etiology and treatment response. The investigators will include individuals with MDD and PDD to provide results that generalize to both conditions and keep the option of exploring each separately.

Objective: The investigators' primary objective is to establish whether a predictive score based on a set of clinical variables and biomarkers differentially predicts response to antidepressant medication and to cognitive-behavioural psychotherapy with a clinically significant effect size. Secondary objectives include testing each predictor separately, testing a gene-environment interaction between the serotonin transporter polymorphism and childhood maltreatment, and integrative analyses using multiple predictors to derive optimized prediction algorithms.

Primary hypothesis: A score reflecting loss of interest and activity, history of childhood maltreatment, systemic inflammation and genetic sensitivity to environment will predict a better response to cognitive-behavioural therapy relative to antidepressant medication among adults with major depressive disorder or persistent depressive disorder. If this hypothesis is supported, the resulting predictor will allow personalized selection of treatment for depression, leading to improved outcomes and healthcare efficiency.

Explanation of changes: An original proposal included a larger sample size (n = 360) across four sites and a genetic variable as one of the predictors. Because the study only got funded at a single site, the sample size will be n = 80, which is insufficient for genetic analyses. Therefore, the genetic predictor was replaced with a measure of metabolism that showed predictive value in a related condition with a larger effect size that indicates that testing may be adequately powered with the reduced sample size.