Data-driven biological subtypes of depression: systematic review of biological approaches to depression subtyping

Abstract

Research into major depressive disorder (MDD) is complicated by population heterogeneity, which has motivated the search for more homogeneous subtypes through data-driven computational methods to identify patterns in data. In addition, data on biological differences could play an important role in identifying clinically useful subtypes. This systematic review aimed to summarize evidence for biological subtypes of MDD from data-driven studies. We undertook a systematic literature search of PubMed, PsycINFO, and Embase (December 2018). We included studies that identified (1) data-driven subtypes of MDD based on biological variables, or (2) data-driven subtypes based on clinical features (e.g., symptom patterns) and validated these with biological variables post-hoc. Twenty-nine publications including 24 separate analyses in 20 unique samples were identified, including a total of ~4000 subjects. Five out of six biochemical studies indicated that there might be depression subtypes with and without disturbed neurotransmitter levels, and one indicated there might be an inflammatory subtype. Seven symptom-based studies identified subtypes, which were mainly determined by severity and by weight gain vs. loss. Two studies compared subtypes based on medication response. These symptom-based subtypes were associated with differences in biomarker profiles and functional connectivity, but results have not sufficiently been replicated. Four out of five neuroimaging studies found evidence for groups with structural and connectivity differences, but results were inconsistent. The single genetic study found a subtype with a distinct pattern of SNPs, but this subtype has not been replicated in an independent test sample. One study combining all aforementioned types of data discovered a subtypes with different levels of functional connectivity, childhood abuse, and treatment response, but the sample size was small. Although the reviewed work provides many leads for future research, the methodological differences across studies and lack of replication preclude definitive conclusions about the existence of clinically useful and generalizable biological subtypes.

Atypical depression shares genetic predisposition with immuno-metabolic traits: a population-based study

“…Clinical heterogeneity may also reflect different underlying biological and causal pathways. Increasing evidence suggests that depressive symptoms and subtypes are differentially associated with genetic risk factors that overlap with other disorders (Beijers, Wardenaar, van Loo, & Schoevers, 2019;Thorp et al, 2019;Milaneschi et al, 2016;Milaneschi et al, 2017a). Investigating the clinical and genetic correlates of more homogenous subtypes may improve the understanding of specific aetiological mechanisms and the development of potential treatment targets.…”

MicroRNA mediators of early life stress vulnerability to depression and suicidal behavior

“…It is still not clear which factors associated with ELS are the most detrimental for later life outcomes or exactly how these events can have such long-lasting effects on brain functions. Furthermore, it has been proposed that depression, subsequent to ELS, comprises a subset of patients with unique etiopathology who would benefit from different treatment or preventative medicine as compared with other patients [28,29]. Further understanding of the molecular neurobiology associated with ELS will bring us closer to identifying vulnerable populations and developing effective treatments.…”

We need an operational framework for heterogeneity in psychiatric research

“…A comprehensive review of these studies is beyond our scope, but many of them have been reviewed elsewhere. 24,25 The central point to appreciate is that these studies all implicitly prioritize multimodality over deviance as the sine qua non of heterogeneity. When we meas-ure heterogeneity by latent cluster counting, we are not interested in the absolute amount of deviation between observations, but rather only in the aggregation of samples into effectively homogeneous groups.…”