Lurasidone (Latuda)

Classification

• Schizophrenia - adults and adolescents aged 13-17 years
• Bipolar I Depression (monotherapy) - adults and pediatric patients aged 10-17 years
• Bipolar I Depression (adjunctive therapy) - with lithium or valproate in adults

Mechanism of Action

Pharmacokinetics

• Bioavailability: 9-19% (low; significantly increased by food)
• Food requirement: Must be taken with at least 350 calories - this is mandatory for adequate absorption
• Tmax: 1-3 hours after dosing
• Volume of distribution: ~6,173 L (extensive tissue distribution)
• Protein binding: High
• Metabolism: Primarily via CYP3A4 - NOT a substrate for CYP1A2, 2D6, 2C9, 2C19, or 2B6
  • Pathways: oxidative N-dealkylation, norbornane ring hydroxylation, S-oxidation
  • Produces 2 active and 2 inactive major metabolites
• Elimination: Feces 80%, urine 9%
• Half-life: ~18 hours (allows once-daily dosing)

Therapeutic Indications and Efficacy

Schizophrenia

• Lurasidone significantly outperformed placebo at 40, 80, and 120 mg/day on BPRSd and PANSS total scores
• One study (N=482) showed lurasidone 80 and 160 mg/day were both superior to placebo, comparable to quetiapine XR 600 mg/day
• One study used olanzapine 15 mg/day as active comparator; both lurasidone and olanzapine significantly beat placebo

Bipolar I Depression - Monotherapy

• 6-week RCT (N=485): Flexible-dose lurasidone 20-60 or 80-120 mg/day vs placebo
• Both dose ranges were superior to placebo on the MADRS score from baseline to week 6
• No significant difference between lower and higher dose ranges

Bipolar I Depression - Adjunctive Therapy

• 6-week RCT (N=340): Patients inadequately controlled on lithium or valproate randomized to lurasidone 20-120 mg/day or placebo
• Lurasidone was statistically superior to placebo in reducing MADRS score as an adjunct to mood stabilizers

Dosing

• No initial dose titration required
• Taken once daily with food (at least 350 kcal)
• Available as 20, 40, 80, and 120 mg tablets

• Moderate-severe renal impairment: Do not exceed 80 mg/day
• Severe hepatic impairment: Do not exceed 40 mg/day
• With moderate CYP3A4 inhibitor: Do not exceed 40 mg/day

Adverse Effects

Common Adverse Reactions

• Somnolence, akathisia, nausea, parkinsonism, agitation

• Akathisia, EPS, somnolence, nausea, vomiting, diarrhea, anxiety

• Akathisia, somnolence, nausea, insomnia, anxiety

Metabolic Profile (Key Advantage)

• Minimal weight gain - mean ~0.43 kg at 6 weeks in schizophrenia trials
• Neutral glucose/lipid effects - no clinically significant changes in cholesterol or fasting glucose
• Low affinity for H1 and M1 receptors explains the favorable metabolic/sedation profile
• Compared favorably to asenapine and iloperidone in metabolic impact

Cardiovascular

• Orthostatic hypotension: ~1.1% (vs 0.9% placebo)
• Does not significantly prolong QTc

Endocrine

• Can increase prolactin levels (as with most antipsychotics)

Cognitive Effects

Contraindications

• Hypersensitivity to lurasidone or any tablet component
• Concomitant use with strong CYP3A4 inhibitors (ketoconazole, clarithromycin, ritonavir, voriconazole, mibefradil)
• Concomitant use with strong CYP3A4 inducers (rifampin, phenytoin, carbamazepine, St. John's wort, avasimibe)

Drug Interactions

Black Box Warnings

1. Increased mortality in elderly patients with dementia-related psychosis - lurasidone, like all SGAs, carries this warning; it is not approved for this population
2. Suicidal thoughts and behaviors in children, adolescents, and young adults - applies due to indication in depression

Other Safety Concerns

• Neuroleptic Malignant Syndrome (NMS): Has been reported; discontinue immediately if suspected
• Tardive Dyskinesia (TD): Risk with all antipsychotics; minimize dose and duration
• Seizures: Use caution in patients with a history of seizures
• Hyperprolactinemia: Can elevate prolactin levels; monitor for galactorrhea, amenorrhea, sexual dysfunction

Comparison to Other Atypicals

Recent Evidence (2023-2026)

• A 2023 Lancet Psychiatry network meta-analysis on pharmacological interventions for acute bipolar depression included lurasidone among effective options
• A 2024 meta-analysis in Psychiatry Research confirmed second-generation antipsychotic effectiveness and safety across psychiatric disorders beyond schizophrenia, including bipolar depression
• A 2026 BJPsych meta-analysis (PMID 40831006) specifically established lurasidone's transdiagnostic efficacy on depressive symptoms across RCTs
• A 2026 network meta-analysis on cognition (PMID 41191868) supports lurasidone's procognitive advantages in schizophrenia-spectrum disorders

Artificial Intelligence in Psychiatry

Overview

1. Historical Roots

2. Core AI Technologies Used in Psychiatry

3. Clinical Applications

3a. Diagnosis and Screening

• Depression and anxiety: NLP tools detect linguistic markers of depression in speech, social media posts, and EHR notes. A 2025 systematic review on NLP for youth mental health prevention (PMID 39986200) found NLP offers precision early identification of at-risk youth
• Schizophrenia: ML models analyze speech disorganization, semantic coherence, and neuroimaging (fMRI, structural MRI) biomarkers. A 2023 narrative review in Consortium Psychiatricum (PMID 38249535) covered ML applications in schizophrenia diagnosis
• Dementia/Alzheimer's: AI analysis of neuroimaging, EEG, and biomarker patterns. A 2023 review in Alzheimer's & Dementia (PMID 37654029) described AI for biomarker discovery across Alzheimer's disease

3b. Suicide and Self-Harm Risk Prediction

• AUC ranged from 0.69 to 0.93
• Sensitivity 45-82%, specificity 91-95%
• Positive predictive value was only 0.1% in low-prevalence populations and 17% in medium-prevalence populations
• Critical conclusion: Current ML algorithms are not accurate enough for population-level suicide screening or treatment allocation. They perform best in high-prevalence clinical settings (e.g., post-discharge psychiatric patients) where PPV reaches ~66%
• The authors recommend needs-based assessment for all patients rather than algorithmic triaging

3c. Treatment Personalization

• Pharmacogenomics + AI: Combining genetic profiles with ML to predict antidepressant/antipsychotic response
• Treatment-resistant depression: ML models identify patients likely to respond to ECT, ketamine, or TMS
• Bipolar disorder: Genomic ML analysis (e.g., 2025 Nature meta-analysis PMID 39843750) generates biological and phenotypic insights that feed treatment stratification

4. Digital Therapeutics and AI-Powered Chatbots

Evidence for AI-CBT Chatbots

5. Large Language Models (LLMs) in Psychiatry

• Suicidal ideation detection from text (38% of studies)
• Conversational mental health agents - accessible, destigmatized eHealth (18% of studies)
• Clinical decision support, psychoeducation, documentation (45% of studies)

• Hallucinations - generating plausible but false clinical information
• Black-box opacity - unclear reasoning limits clinician trust
• No benchmarked ethical framework
• Data privacy - sensitive mental health data handled by commercial systems
• Overreliance risk - patients and clinicians delegating too much to AI

• Generating clinical documentation drafts (AI scribe tools)
• Patient psychoeducation
• Symptom triage and onboarding
• Training and simulation for psychiatric trainees

6. Neuroimaging and Biomarker Analysis

• AI analysis of fMRI, structural MRI, and PET scans identifies neural circuit abnormalities in depression, schizophrenia, PTSD, and OCD
• EEG-based ML models analyze brainwave patterns for depression subtypes and predict ECT response
• Deep learning applied to retinal imaging is emerging as a low-cost proxy biomarker for early neurodegenerative psychiatric conditions

7. Administrative and Workflow Applications

• AI scribes: Automatically transcribe therapy/psychiatry sessions and generate SOAP notes, reducing documentation burden. One in four therapists in the US now use AI in practice
• Scheduling optimization: Predictive tools reduce no-show rates, optimize appointment slots
• Prior authorization: NLP automates insurance documentation
• Coding and billing: AI-assisted ICD/CPT code assignment from clinical notes

8. Ethical and Safety Challenges

9. Current Limitations

• Most AI psychiatric tools validated on narrow, non-diverse research cohorts
• Generalizability across cultures, languages, and clinical settings remains poor
• Most evidence comes from short-term trials; long-term outcome data is scarce
• Real-world clinical integration (EHR connectivity, workflow embedding) lags behind research
• Regulatory frameworks (FDA, CE marking) for AI-as-medical-device are still maturing

10. Future Directions (2025-2030)

• Multimodal AI: Combining voice tone, facial expression, language, physiology (wearables), and neuroimaging for richer psychiatric assessment
• Federated learning: Training AI on distributed hospital datasets without sharing raw patient data - addresses privacy while improving model robustness
• Precision psychiatry: AI-guided matching of patients to the right drug, dose, or therapy type from the first visit
• VR + AI: Immersive therapeutic environments for PTSD, phobias, and social anxiety, guided by AI feedback loops
• Longitudinal monitoring: Passive sensing via smartphone (screen time, typing patterns, GPS data) to detect early relapse signals
• Drug repurposing: ML across GWAS data identifies existing drugs as new psychiatric treatments (e.g., 2026 schizophrenia genomics meta-analysis PMID 42421611)

Summary Table