Emotional hyper-reactivity and cardiometabolic risk in remitted bipolar patients: a machine learning approach

Acta Psychiatr Scand. 2018 Oct;138(4):348-359. doi: 10.1111/acps.12901. Epub 2018 May 15.


Objective: Remitted bipolar disorder (BD) patients frequently present with chronic mood instability and emotional hyper-reactivity, associated with poor psychosocial functioning and low-grade inflammation. We investigated emotional hyper-reactivity as a dimension for characterization of remitted BD patients, and clinical and biological factors for identifying those with and without emotional hyper-reactivity.

Method: A total of 635 adult remitted BD patients, evaluated in the French Network of Bipolar Expert Centers from 2010-2015, were assessed for emotional reactivity using the Multidimensional Assessment of Thymic States. Machine learning algorithms were used on clinical and biological variables to enhance characterization of patients.

Results: After adjustment, patients with emotional hyper-reactivity (n = 306) had significantly higher levels of systolic and diastolic blood pressure (P < 1.0 × 10-8 ), high-sensitivity C-reactive protein (P < 1.0 × 10-8 ), fasting glucose (P < 2.23 × 10-6 ), glycated hemoglobin (P = 0.0008) and suicide attempts (P = 1.4 × 10-8 ). Using models of combined clinical and biological factors for distinguishing BD patients with and without emotional hyper-reactivity, the strongest predictors were: systolic and diastolic blood pressure, fasting glucose, C-reactive protein and number of suicide attempts. This predictive model identified patients with emotional hyper-reactivity with 84.9% accuracy.

Conclusion: The assessment of emotional hyper-reactivity in remitted BD patients is clinically relevant, particularly for identifying those at higher risk of cardiometabolic dysfunction, chronic inflammation, and suicide.

Keywords: C-reactive protein; bipolar disorder; cardiometabolic dysfunction; emotional hyper-reactivity; machine learning.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Adult
  • Affective Symptoms* / blood
  • Affective Symptoms* / epidemiology
  • Affective Symptoms* / etiology
  • Affective Symptoms* / physiopathology
  • Bipolar Disorder* / blood
  • Bipolar Disorder* / complications
  • Bipolar Disorder* / epidemiology
  • Bipolar Disorder* / physiopathology
  • Blood Glucose
  • Blood Pressure / physiology
  • Cardiovascular Diseases* / epidemiology
  • Cardiovascular Diseases* / physiopathology
  • Female
  • France / epidemiology
  • Glucose Metabolism Disorders* / blood
  • Glucose Metabolism Disorders* / epidemiology
  • Glycated Hemoglobin A
  • Humans
  • Machine Learning*
  • Male
  • Middle Aged
  • Remission Induction
  • Risk
  • Suicide, Attempted / statistics & numerical data*


  • Blood Glucose
  • Glycated Hemoglobin A
  • hemoglobin A1c protein, human