From the inside-out: Using Latent Biomarker Profiles to Identify Distinct Depressive Symptom Networks in People with Major Depression

Objective: Major depression is a common mental health problem that is difficult to treat, despite the wide array of evidence-based treatments. To improve treatment outcomes for those that experience major depression, several studies have identified latent classification of groups of persons who experience depression. Several recent studies have attempted latent grouping based on biomarkers. However, studies have been inconsistent in biomarker selection, tend to use low quality measures, and do not discuss symptom level differences across subtypes. Thus, the present study aimed to use latent profile analysis to measure biological dysregulation among individuals experiencing major depression using a previously established measure and then identify how symptoms may differ across groupings

Method: Secondary data from two waves of the Midlife Development in the United States study was used, including a sample of 489 individuals experiencing a current major depressive episode. Latent profile analysis of 17 biomarkers across 7 different biological systems (sympathetic nervous system, peripheral nervous system, hypothalamic-pituitary-adrenal axis, inflammation, cardiovascular, glucose, lipids), was used to identify latent biological profiles. Selection of biomarkers was based on a previous establishment of allostatic load factor structure, a broad model of biological dysregulation (Wiley et al., 2016). Number of latent profiles was based on model fit indices (AIC, BIC, Entropy) and overall parsimony. A psychometric network analysis was then estimated for each group with items of the Centers for Epidemiologic Studies Depression Scale (CES-D) serving as nodes. Strength centrality (SC) was then used to compare the influence of nodes in each model.

Results: Average age of the sample was about 50 years. About 74% of participants identified as non-Hispanic White, and 43% identified as male. Mean CES-D scores for the sample were 21.66 out of 60. Latent profile analysis identified a five-profile solution: High Peripheral Nervous System (n=73); Low Inflammation & Mixed Lipids (n=86); Low Peripheral Nervous System & High Glucose (n=70); Low Sympathetic Nervous System & Low Cortisol (n=159); High Peripheral Nervous System & Low Pulse (n=101). No mean differences in depressive symptoms emerged across profiles (p=0.83). However, network analysis identified significant differences in depressive symptoms across models. For instance, High Peripheral Nervous System & Low Pulse had high SC for endorsing feelings of depression and sadness, while Low Peripheral Nervous System & High Glucose had higher SC for denying feelings of happiness and enjoyment. Low Sympathetic Nervous System & Low Cortisol had higher SC related to interpersonal difficulty, while Low Peripheral Nervous System & High Glucose had higher SC related to psychomotor disturbance.

Conclusion: The present study demonstrates that latent profile analysis of biomarkers can yield biological subtypes that are associated with distinct depressive symptom typology. Future intervention efforts that focus on a whole-health approach to major depression treatment may consider how biological and psychological mechanisms should be dually targeted. Future research should aim to replicate these findings with a larger and more racially/ethnically representative sample, as well as across different depressive symptom measures.