Using Social Risk Factors to Predict Allostatic Biotypes of Depression: A Latent Profile and Multinomial Regression Analysis

Background and Purpose: Major depression is a common but challenging condition to treat. Certain psychosocial risk factors such as childhood maltreatment, loneliness, and recent stress are linked to more severe symptoms and greater likelihood of treatment resistance. These experiences may influence inflammation and contribute to broader biological dysregulation, as suggested by allostatic load theory. This study examines how these psychosocial risks relate to distinct biological profiles, with the goal of informing new treatment approaches for depression.

Methods: We used data from the Midlife in the United States study (MIDUS), drawing a sample of 367 participants currently experiencing depression based on CES-D scores of 16 or greater. Then, using latent profile analysis, we created allostatic profiles of individuals currently experiencing depression, using a total of 22 biomarkers across 7 biological systems: inflammation, glucose, sympathetic nervous system (SNS), parasympathetic nervous system (PNS), hypothalamic-pituitary-adrenal (HPA)-axis, cardiovascular, and lipids. After identifying allostatic profile solutions, we engaged in a model building strategy amenable to our sample size through a multinomial regression framework to understand differences in predictors of profile membership. The first model evaluated common risk factors for depression including childhood maltreatment, loneliness, and recent perceived stress. The second model evaluated health factors including the typical number of alcoholic drinks consumed, current tobacco smoking status, and regular engagement in exercise. The third model evaluated potential demographic differences including self-reported sex, racial/ethnic identity, age, and education. We then tested a final multinomial regression model, including only significant predictors identified through the first three models.

Results: A 5-profile solution was identified as the best fit to the data: High Inflammation and Glucose (n = 74), Low PNS (n = 83), High PNS (n = 73), Low Inflammation & High HDL (n = 67), and High HPA-Axis (n = 70). We selected the Low Inflammation & High HDL profile as the reference group for multinomial regression modeling. The final model showed that childhood maltreatment was associated with an increased risk for membership in the High Inflammation & Glucose (b = 0.033 SE = 0.011, p < 0.01) and the Low PNS (b = 0.031, SE = 0.011, p <0.01) groups relative to the Low Inflammation & High HDL group. Neither loneliness nor recent perceived stress were significant predictors of group membership. Childhood maltreatment continued to predict membership of the High Inflammation & Glucose and Low PNS groups after adjusting for smoking status, exercise, racial/ethnic identity, age, and education.

Conclusions and Implications: Experiences of childhood maltreatment, but not loneliness or recent stress, differentially predicts biological profiles of depression. Given that childhood maltreatment is a risk factor for an overall worse course of illness, improving outcomes for people who experience depression and childhood maltreatment may be to identify treatments that differently target inflammation, glucose, and PNS. Cognitive behavioral therapy, collaborative care models for diabetes and depression, anti-inflammatory medications, and vagus nerve stimulation can be used to target these systems and may be important treatment considerations specifically for those with a history of childhood maltreatment.