Towards personalised antidepressive medicine based on "big data": an up-to-date review on robust factors affecting treatment response.

Prescribing antidepressant medication is currently the most effective way of treating major depression, but only very few patients achieve permanent improvement. Therefore, it is important to identify objectively measurable markers for effective, personalized therapy. The aim of this review article is to collect all the markers that are robustly predictive of the outcome of therapy. We searched for systematic review articles that have simultaneously investigated the effects of as many different markers as possible on the response to antidepressant therapy in major depressive patients. From these we extracted markers that have been found to be significant by at least two independent review studies and that have proven replicable also within each of these reviews. A separate search was performed for meta-analyses of pharmacogenetic genome-wide association studies. Based on our results, onset time, symptom severity, presence of anhedonia, early treatment response, comorbid anxiety, alcohol consumption, frontal EEG theta activity, hippocampal volume, activity of anterior cingulate cortex, as well as a peripheral marker, serum BDNF levels have proven replicable predictors of antidepressant response. Pharmacogenomic studies to date have not yielded replicable results. Predictors identified as robust by our study may provide a starting point for future machine learning models within a 'big data' database of major depressive patients. (Neuropsychopharmacol Hung 2022; 24(1): 17-28).