Can we do a better job of antidepressant selection based on symptom clusters, analyzed using machine learning techniques?
Perhaps so. And if so we will avoid a great deal of unnecessary suffering as people try one medication after another until the find the right medication or combination of medications to treat their depression.
That is the premise behind an ambitious initiative developed by some Yale researchers, summarized in an article in JAMA Psychiatry in February 2017.
The researchers first used data collected from the largest study of antidepressant treatment selection to date, the STAR*D study which included 4039 patients sequentially treated with antidepressants until they achieved a remission of symptoms.
From this initial set they identified certain “clusters” of symptoms that seemed to be related to each other.
In the figure at the right this initial symptom cluster is listed under “A”. It comprises three sets of symptoms from the QIDS-SR – a standard depression rating scale that we use often at Gateway to measure clinical progress.
The symptom clusters were
- Insomnia – Symptoms based on questions about difficulty getting to sleep (sleep onset insomnia), staying asleep (midnocturnal insomnia) or waking too early (early morning insomnia).
- Core Emotional – Symptoms of fatigue or lack of energy, loss of interest or motivation, difficulty concentrating or making decisions, feeling sad blue or down, and feeling self-critical or worthless.
- Atypical – A set of symptoms that might better have been labelled “other” – psychomotor agitation or restlessness, psychomotor slowing (feeling slowed down in thought, speech, or movement), suicidal thoughts, and increased sleeping.
The took this list of symptoms and sought to determine whether they could replicate the symptom clusters in data from another large clinical trial (CO-MED) and they also looked at another clinical instrument that was part of the STAR-D trial, the clinician rated Hamilton Depression rating scale see if they could identify the symptom clusters using that instrument.
They found significant evidence that these three symptom clusters could be found to occur in all three sets of data.
The next question was to look at the outcomes data to see if different medications had better effects depending on the most prominent symptoms that the patient presented with.
“Combined escitalopram and bupropion treatment was significantly more effective in treating core emotional symptoms than citalopram (ES, 0.7 QIDS-SR points; 95%CI, 0.2 to 1.3; P = .03). For sleep/insomnia symptoms, venlafaxine with mirtazapine outperformed citalopram (ES, 1.4; 95%CI, 1.0to 1.8; P < .001). For core emotional symptoms in HAM-D scale trials (Figure 2B), high-dose duloxetine outperformed escitalopram (ES, 2.3 HAM-D points; 95%CI, 1.6 to 3.1; P < .001). Escitalopram was not significantly different from placebo for core emotional symptoms (ES, 0.03 HAM-D points; 95% CI, −0.7 to 0.8; P = .94). For sleep symptoms, high-dose duloxetine outperformed fluoxetine (ES, 0.9; 95% CI, 0.1 to 1.7; P = .046). For atypical symptoms,high-dose duloxetine outperformed all others (ES,0.5-1.9) and escitalopram was worse than placebo (ES, 0.7; 95% CI, 0.3 to 1.1; P = .002). Among our HAM-D studies, only 2 antidepressant treatments (high-dose duloxetine and paroxetine) outperformed placebo for all 3 symptom clusters.”
If we break these results down using what we understand about the pharmacology of these medications some possible principles can be identified.
- Core Emotional Symptoms, particularly on the clinician rated HAM-D instrument, are minimally improved by SSRI’s alone (citalopram or escitalopram). Combined escitalopram and bupropion and high dose duloxetine (120 mg) were significantly more effective than SSRI’s.
- For Sleep Symptoms venlafaxine plus mirtazapine or high-dose duloxetine (SNRI’s) outperformed SSRI’s (fluoxetine or escitalopram).
- Atypical Symptoms responded best to high dose duloxetine (SNRI), and escitalopram (SSRI) was worse than placebo.
Given the overall findings in this study, it remains somewhat of a mystery to me why it has been so hard to find an overall benefit to using SNRI’s over SSRI’s in treating depression. But these results do fit with my own clinical impressions and my practice which is to use SSRI’s for treating anxiety and anxious depression more than core depressive symptoms.
We will be using the instrument developed by this group as part of our assessment and treatment selection process for patients with unipolar depression.
Chekroud AM, Gueorguieva R, Krumholz HM, Trivedi MH, Krystal JH, McCarthy G. Reevaluating the Efficacy and Predictability of Antidepressant Treatments – A Symptom Clustering Approach. JAMA Psychiatry. Published online February 22, 2017. doi:10.1001/jamapsychiatry.2017.0025