Abstract by Lukas Frederic Westphal-Ostrowski

Selecting the best antidepressant for a patient remains challenging. Current treatment guidelines provide limited support for prescribing decisions, as the average treatment effects of second-generation antidepressants are broadly similar across agents. To improve outcomes, evidence that accounts for patients' baseline characteristics and personal preferences is needed to guide the choice of an agent with the most favorable balance between effectiveness and adverse effects. However, the vast and fragmented literature on antidepressants makes it difficult for clinicians and researchers to navigate existing evidence. In this context, this thesis aims to advance both the methodological toolkit and the evidence base supporting antidepressant prescribing through three complementary studies at progressively narrowing analytical scales.

Study I applied network analysis with unsupervised graph-based clustering to map 38,961 publications related to selective serotonin reuptake inhibitors (SSRIs) from 1982 to 2025, identifying thematic evolution and fragmentation in the research field.
Study II employed Danish national registries and a new-user, active-comparator design to compare the risk of sexual dysfunction between users of four commonly prescribed antidepressants.
Study III developed and evaluated an individualized treatment rule to reduce treatment non-response in 17,645 antidepressant-naive patients hospitalized for depression using a target trial emulation. 

Study I revealed increasing fragmentation of SSRI research over time and enabled the development of an interactive map offering an intuitive and inclusive way to explore the extensive research corpus.
Study II demonstrated clinically meaningful differences in sexual dysfunction risk: venlafaxine was associated with a 27\% higher risk compared to citalopram, mirtazapine showed modest evidence of lower risk, and sertraline did not differ significantly from citalopram.
Study III showed that an individualized treatment rule reduced the absolute risk of treatment non-response by 5.47\% compared with universal sertraline prescription, representing the current standard of care.

This thesis yields two complementary contributions. First, it provides a navigation tool for exploring the antidepressant literature and supporting more targeted research questions using similar methodology. Second, it demonstrates that routinely collected healthcare data can support evidence-based treatment selection at both population and individual levels. While antidepressants exhibit similar average effects, their adverse event profiles differ meaningfully, and treatment effect heterogeneity can be leveraged to improve outcomes through individualized treatment rules. Future research should integrate multiple outcomes simultaneously within individualized prediction frameworks to better reflect the complex trade-offs inherent in antidepressant selection and move toward patient-centered treatment recommendations.