Modelling PTSD diagnosis using sleep, memory, and adrenergic metabolites: An exploratory machine-learning study

Area of Science:

• Neuroscience
• Psychiatry
• Sleep Medicine

Background:

• Posttraumatic stress disorder (PTSD) is characterized by sleep disturbances, impaired declarative memory, and hyperarousal.
• Understanding the specific pathophysiological changes associated with PTSD is crucial for improved diagnosis and treatment.

Purpose of the Study:

• To evaluate if combined features of sleep disturbances, impaired declarative memory, and hyperarousal can delineate pathophysiological changes in PTSD.
• To determine the diagnostic utility of sleep, cognitive, and biochemical variables in distinguishing PTSD from trauma exposure without PTSD.

Main Methods:

• Recruited three groups: PTSD-diagnosed individuals (N=20), trauma survivors without PTSD (TE; N=20), and healthy controls (HC; N=20).
• Utilized between-group difference analyses and support vector machine (SVM)-learning on participant features, including subjective/objective sleep, declarative memory, and metabolite data.

Main Results:

• PTSD individuals reported poorer sleep quality, objectively showed less sleep depth, and had declarative memory deficits compared to HC.
• SVM-learning distinguished HC from trauma participants with 80% accuracy using sleep and memory features.
• PTSD and TE groups were differentiated with 70% accuracy based on sleep variables alone.

Conclusions:

• Sleep characteristics are primary features differentiating individuals with PTSD from those without.
• Exploratory SVM-learning analysis provides a framework for future PTSD research focused on sleep and memory.
• Findings suggest potential for improving PTSD diagnostic accuracy and treatment through sleep- and memory-based investigations.