Depression Symptom Clusters May Predict Treatment Response

February 27, 2017

By Marilynn Larkin

NEW YORK—Depression symptoms can be grouped into three clusters that can be used to predict a patient’s response to antidepressants, researchers suggest.

Adam Chekroud of Yale University told Reuters Health, “We know that depression includes a wide range of symptoms, from low mood and feeling worthless to problems sleeping, slowed thinking, and suicidal ideation. We wanted to know whether antidepressants work well in treating all of these symptoms, or whether they are primarily effective on certain kinds of symptoms.”

To investigate, Chekroud and colleagues first analyzed data from a multicenter trial that compared various combinations of depression treatments in 4,706 patients (about one-third men, mean age 43). They identified three clusters of symptoms at baseline, based on responses to the self-report Quick Inventory of Depressive Symptomatology (QIDS) scale.

“The clusters were a group of mood/emotional symptoms, a group of insomnia symptoms, and a group of ‘atypical’ symptoms, like psychomotor symptoms or suicidal ideation, that were a bit less common,” Chekroud explained by email.

The team replicated the findings in a group of 640 patients with similar demographics, using the clinician-rated Hamilton Depression (HAM-D) rating scale. “This gave us some confidence that the grouping of symptoms was meaningful,” he noted. “Then we could return to our original question - whether antidepressants work well on all symptoms or just some.”

To see whether symptom clusters were equally responsive to different antidepressants and whether certain drugs or doses were more effective than others, the team used predictive machine-learning algorithms to analyze outcomes from nine clinical trials including 7,221 patients.

“In general, antidepressants were very effective in treating mood and emotional symptoms like low mood, feeling worthless, or excessive worrying,” Chekroud observed. “Some antidepressants were also good at treating a group of sleep symptoms. However, antidepressants were far less effective in treating the third cluster of symptoms.”

As reported in JAMA Psychiatry, online February 22, no antidepressant worked equally well across all three symptom clusters. For example, for citalopram, escitalopram with placebo, and escitalopram with bupropion, trajectories were significantly better for emotional symptoms than for either sleep or atypical symptoms when measured according to the QIDSSR (all P<0.001). Sleep trajectories were also better than atypical trajectories for these specific treatments (all P

Efficacy differences were often greater between drugs than between treatments and placebo. For example, high-dose duloxetine outperformed escitalopram in treating core emotional symptoms (2.3 points improvement in HAM-D scores over two months; P<0.001). However, outcomes with escitalopram were not significantly different from placebo (0.03 improvement in HAM-D points; P=0.94).

“It is informative to think about depression in terms of clusters of symptoms, and we can tailor our treatment selection towards each patient’s specific symptom profile,” Chekroud concludes.

Dr. David Hellerstein, a professor of clinical psychiatry at Columbia University Medical Center in New York City, called the study “impressive.” He told Reuters Health that the machine learning approach used in the study “is being used to identify patterns related to better response to treatment across many branches of medicine, including psychiatry.”

“Machine learning finds patterns to predict treatment response in the data itself, rather than relying on preconceptions of researchers or clinicians about which symptoms are most important or how they are interrelated,” he explained by email. Since most researchers believe that major depression is heterogeneous, “there’s a significant value if it’s possible to disentangle subtypes, both for medicine response and for understanding the differing biology of these presumably different conditions.”

The findings might help personalize treatment, aid in the development of new drugs that focus on specific cluster symptoms and “could possibly guide studies of the biology of depression since different symptom clusters may reflect different abnormalities in the brains of people with depression,” he said.

Dr. Hellerstein noted several limitations of the analysis, including different study designs, some variation in symptom clusters and the focus on medication studies. “Some forms of psychotherapy (CBT, behavioral activation) may be useful for symptoms that don’t respond well to medicines,” he observed.

“The biggest limitation to me as a practicing psychiatrist is that it’s not clear how well these findings can be used in the care of individual patients - for one thing, studies often exclude people with medical or other problems like substance abuse, etc.,” he said.

And, he wonders, how many patients would he need to treat according to these predictions to get one additional patient better, compared to choosing an antidepressant at random?

A patient self-assessment tool based on the study is posted at www.spring.care/spring-assessment.

Chekroud holds equity in Spring Care Inc, a behavioral health startup developing digital tools to improve the treatment of mental illness. He is also inventor on a patent by Yale University about treatment selection for depression.

SOURCE: http://bit.ly/2lNWOZ2

JAMA Psychiatry 2017.

(c) Copyright Thomson Reuters 2017. Click For Restrictions - http://about.reuters.com/fulllegal.asp