Pharmaco-multiomics in major depressive disorder: a narrative review and proposed translational framework for difficult-to-treat and treatment-resistant depression

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Background Antidepressant prescribing in major depressive disorder (MDD) remains dominated by sequential trial and error, particularly when illness becomes difficult to treat (DTD) or meets conventional treatment-resistant depression (TRD) criteria. Objective This narrative review evaluates pharmacogenomics (PGx), therapeutic drug monitoring (TDM), metabolomics/lipidomics, immune-inflammatory markers, proteomics, transcriptomics, epigenomics, microbiomics, and multi-omics machine learning according to their capacity to improve defined pharmacological decisions. Methods Evidence was identified through structured PubMed/MEDLINE searches updated to 29 July 2026, targeted searches of ClinicalTrials.gov and the WHO International Clinical Trials Registry Platform, and backward and forward citation searching. The review was restricted to adult MDD and used author-assigned categories that distinguish prognostic, predictive, pharmacokinetic, and treatment-emergent monitoring markers and differentiate discovery, independent replication, external validation, and clinical utility. The article is a narrative Review, not a PRISMA systematic review; study selection, extraction, and translational classification were performed by one author without formal study-level risk-of-bias grading. Results Most evidence derives from general adult MDD rather than prospectively defined DTD/TRD cohorts. Drug-specific, guideline-supported PGx and indication-specific TDM are the most clinically proximal molecular tools, but their value is bounded and depends on medication history, inhibitors and inducers, adherence, organ function, assay coverage, and phenoconversion. Commercial combinatorial PGx trials show small and sometimes nonpersistent clinical effects. Metabolomic, inflammatory, proteomic, transcriptomic, epigenomic, microbiomic, and integrated machine-learning studies identify plausible mechanisms and candidate predictors, but the majority remain discovery-stage or internally validated. Positive subgroup signals, including C-reactive protein-defined differential response, are counterbalanced by null prospective trials and inconsistent thresholds. Conclusion Current clinical application should remain limited to systematic medication and interaction review, drug-specific guideline-supported PGx, and indication-specific TDM. Dynamic omics layers should be treated as investigational until locked models are externally validated and prospective biomarker-guided strategies demonstrate incremental decision value, clinical benefit, feasibility, equity, and cost-effectiveness. The review proposes a future-state architecture and staged evidence-to-implementation framework for DTD/TRD.