Andrew D. Krystal

Approach–avoidance refers to the psychological conflict in which deciding to approach a reward also carries potential punishment. The orbitofrontal cortex (OFC) is proposed to arbitrate approach–avoidance decisions, yet little is known about the real-time computations within OFC during these decisions. We conducted intracranial recordings in humans implanted with stereotactic electroencephalography electrodes in the OFC while they played a gamified approach–avoidance task. Here we identified two anatomically distinct signals within the OFC prior to approach: an increase in activity within the medial compartment and a decrease within the lateral compartment. Prior to decisions, these signals rapidly alternated between discrete pro-approach and pro-avoidant states, thus introducing a new functional architecture within the human OFC during decision-making.

Objective: Racial sleep disparities may lead to long-term cognitive health consequences, with a specific impact on memory formation and maintenance. Few studies have examined the physiological sleep features, including nonrapid eye movement (NREM) stage 2 and 3 sleep and the corresponding electroencephalographic activity indicative of these stages (ie, delta oscillations 0.5 to 4 Hz), that are sensitive to racial stress and result in memory vulnerability, which is the goal of this investigation. Methods: We assessed self-report and polysomnographic indices of sleep, psychosocial outcomes, and memory performance in a sample of young, Black ( n =63) and White ( n =77) adults living in the United States. Results: Black participants self-reported more experiences of discrimination, lower incomes, perceived social status, and educational attainment, as well as worse subjective sleep quality. Racial group membership was associated with differential presleep memory profiles for both neutral and negatively valenced word pairs; however, memory disparities were attenuated after accounting for sociodemographic and psychosocial factors. Overnight sleep architectural differences emerged, with Black participants spending more time in stage 2 sleep, less time in stage 3 sleep, and having reduced NREM delta power when compared with White participants. NREM delta activity was a significant mediator of racial group membership and postsleep memory performance, especially for negative valenced word pairs. Conclusion: These findings suggest that known racial sleep disparities likely lead to daily, measurable deficits in memory for Black Americans and that NREM delta activity may be a uniquely sensitive biopsychosocial factor that can support cognitive resilience for those experiencing racialized contexts.

Background: Large language models (LLMs) and, more recently, large reasoning models (LRMs) have rapidly garnered significant interest for application in psychiatry and behavioral health. However, recent studies have identified significant shortcomings and potential risks in the performance of LLM-based systems, complicating their application to psychiatric diagnosis. Two promising approaches to addressing these challenges and improving the efficacy of these models are simulated reasoning (SR) and self-verification (SV), in which additional “reasoning tokens” are used to guide model output, either during or after inference. Objective: We aimed to explore how the use of SR (via LRMs) and SV (via supplemental prompting) affects the psychiatric diagnostic performance of LLMs. Methods: 106 case vignettes and associated diagnoses were extracted from the DSM-5-TR (Diagnostic and Statistical Manual, Version 5, Text Revision) Clinical Cases book, with permission. Both an LLM and an LRM model were selected from the latest available model generation for each of the two vendors studied (OpenAI and Google). Two inference approaches were developed: a Basic approach that directly prompted models to provide diagnoses and a SV approach that augmented the Basic approach with additional prompts. All case vignettes were processed by the two LLMs, two LRMs, and two inference approaches, and diagnostic performance was evaluated using the sensitivity and positive predictive value (PPV). Binomial generalized linear mixed models were used to test for significant differences between the model vendors (OpenAI, Google), type (LLM, LRM), and the addition of an SV prompt. Results: All vignettes were successfully processed by each model and inference approach. Sensitivity ranged from 0.732 to 0.817, and PPV ranged from 0.534 to 0.779. The best overall performance was found in the o3-pro LRM using SV, with a sensitivity of 0.782 and a PPV of 0.779. No statistically significant fixed effects were found for sensitivity. For PPV, a statistically significant effect was found for prompt type (SV, P =.007) and model type (LRM, P =.009). No significant interaction effects were identified. Conclusions: We found that both SR and SV yielded statistically significant improvements in the PPV, without significant differences in the sensitivity. The addition of the manually specified SV prompt improved the PPV even when simulated reasoning was used. This suggests that future efforts to apply language models in behavioral health could benefit from manually crafted reasoning prompts and automated SR.

Background: Insomnia during pregnancy is a modifiable risk factor for depression. Research in nonpregnant populations and preliminary findings in a pregnant population suggest that targeting insomnia with digital cognitive behavioral therapy for insomnia (CBT-I) may prevent depression. Objective: This study aims to evaluate the efficacy of digital CBT-I for the prevention of depression through 12 months post partum compared to that of a credible, clinically relevant comparator intervention, sleep hygiene education (SHE). This project will also provide valuable information about how and for whom CBT-I prevents perinatal depression. Methods: The Perinatal Research for Improving Sleep and Mental Health study is a fully remote, 2-arm parallel-group randomized controlled trial designed to test the superiority of digital CBT-I versus SHE for the prevention of perinatal depression. We aimed to randomize a target sample of at least 443 pregnant participants with insomnia disorder to digital CBT-I or SHE at a 1:1 ratio. Participants randomized to digital CBT-I received SleepioRx (Big Health, Inc), which uses sleep consolidation, stimulus control, cognitive therapy, relaxation techniques, and SHE in 6 interactive modules (5‐15 min each). Participants randomized to SHE received information about sleep hygiene, the impact of sleep on performance, healthy sleep habits, lifestyle influences on sleep, and creating a sleep-friendly bedroom in 6 weekly digital handouts. Data are collected at baseline, 5 weeks after randomization, 10 weeks after randomization, 36 weeks’ gestation, and 3, 6, 9, and 12 months post partum. The primary outcome is depression incidence, measured using the Structured Clinical Interview for DSM-5 ( Diagnostic and Statistical Manual of Mental Disorders [Fifth Edition]) Disorders, Research Version. Other outcome measures include the Edinburgh Postnatal Depression Scale, Patient Health Questionnaire-9, Columbia Suicide Severity Rating Scale, Generalized Anxiety Disorder Scale, and a birth outcomes questionnaire. Mediator measures include the Insomnia Severity Index, Consensus Sleep Diary, Structured Clinical Interview for Sleep Disorders, Perseverative Thinking Questionnaire, Dysfunctional Beliefs about Sleep Scale, Pre-sleep Arousal Scale, and Difficulties in Emotion Regulation-Short Form. Results: Data collection began in November 2022 and is projected to end in December 2026. Conclusions: If study hypotheses are supported, then this project will provide a new, scalable approach for preventing perinatal depression.

Background: Cognitive behavioral therapy (CBT) is recommended as the first-line treatment for insomnia; however, few patients have access to it. A new class of Food and Drug Administration (FDA)–regulated digital CBT treatments has the potential to address this unmet need. These treatments are ordered or prescribed by health care providers and are fully automated, delivering CBT directly to patients without human coaches. This trial builds upon promising earlier digital cognitive behavioral therapy for insomnia (CBT-I) research by using a decentralized design to recruit a sample with greater representation of the US general population, including individuals from lower socioeconomic status groups who often face greater barriers to care. Objective: This decentralized trial evaluated the effectiveness of a fully automated digital CBT-I program (SleepioRx) for treating insomnia disorder compared with online sleep hygiene education (SHE) in a sample of participants recruited from across the United States. Methods: A decentralized, parallel-group randomized controlled trial was conducted between November 2022 and August 2023. Participants were recruited nationally from across the United States, and a total of 336 adults aged 22 and older, diagnosed with the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) insomnia disorder via structured clinical interview, were allocated 1:1 to either digital CBT-I (SleepioRx) or online SHE. The primary end points were insomnia severity, assessed using the Insomnia Severity Index (ISI), and sleep diary measures of sleep onset latency (SOL) and wake after sleep onset (WASO) at 10 weeks, with follow-up assessments at 16 and 24 weeks postrandomization. Results: Compared with SHE, SleepioRx showed statistically and clinically significant improvements on the ISI at posttreatment (10 weeks; Cohen d=0.60, P<.001), with effects sustained at follow-up (16 weeks; d=0.65, P<.001; and 24 weeks, d=0.77, P<.001). SleepioRx led to significant reductions in WASO at all time points (10 weeks, P=.003; 16 and 24 weeks, P<.001); however, effects on SOL were not statistically significant at an adjusted α (10 weeks, P=.01; 16 weeks, P=.07; 24 weeks, P=.27). SleepioRx participants had 2.5 times (odds ratio 2.52; P<.001, 99% CI 1.33-4.75) and 5.8 times (odds ratio 5.78; P<.001, 99% CI 2.11-15.84) greater odds of response and remission at week 10, respectively, with statistically and clinically significant differences in rates sustained at follow-up assessments (P<.001). SleepioRx also demonstrated sustained improvements in secondary sleep and broader mental health outcomes. Conclusions: The results of this trial demonstrate the effectiveness of digital CBT-I (SleepioRx) for treating insomnia, with gains sustained at 6 months, and support the FDA authorization of SleepioRx for the treatment of insomnia disorder. These findings underscore the potential of a new class of FDA-authorized, fully automated digital treatments to provide first-line, guideline-recommended CBT at scale. Efforts should now focus on expanding access to these evidence-based treatments. Trial Registration: ClinicalTrials.gov NCT05541055; https://clinicaltrials.gov/ct2/show/NCT05541055

Deep brain stimulation has been used to treat severe, refractory obsessive-compulsive disorder (OCD) with variable outcomes across multiple anatomical targets. To overcome these limitations, we developed an invasive brain mapping paradigm in which electrodes were implanted across the OCD cortico-striato-thalamo-cortical circuit. We then performed extensive stimulation mapping during a multi-day inpatient stay to identify personalized therapeutic targets and characterize their downstream circuit effects. We found two targets within the right ventral capsule (VC) that acutely reduced OCD symptoms. Prolonged VC stimulation suppressed high frequency activity within the structurally and functionally connected orbitofrontal and cingulate cortex, which were identified to be cortical nodes encoding the severity of OCD symptoms. These VC sites were implanted for DBS and combined stimulation of these targets led to a rapid therapeutic response. This case provides the first proof-of-concept that invasive brain mapping can be used to guide a novel personalized, multi-site neuromodulation approach to treat refractory OCD.