Escitalopram treatment for patients with major depressive disorder: decision trees for treatment algorithm

Current treatment algorithms for major depressive disorder (MDD) lack dynamic prediction capabilities, leading to delayed therapeutic adjustments. This study sought to develop escitalopram-specific decision tree models to identify critical treatment adjustment time points and optimize personalized treatment strategies for MDD.

Using longitudinal data from two multicenter studies in China (2015-2020), we analyzed 800 patients with MDD receiving escitalopram monotherapy. Decision tree models incorporated baseline characteristics (age, BMI, disease duration, depressive symptoms) and dynamic treatment parameters (dose, 2-/4-week improvement) to predict full response (>50% symptom reduction) or non-full response (≤50% reduction) at weeks 2 and 4, and remission status (QIDS-SR16≤5 vs. >5) at week 8. Model performance was assessed by accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the curve (AUC).

The week 2 model (n = 800) identified BMI, age, disease duration, course and baseline symptom severity as primary predictors (accuracy = 61.88%, NPV = 84.04%). By week 4 (n = 650), early response status (week 2) merged as a key predictor (accuracy = 69.23%, NPV = 71.62%). The week 8 model (n = 456) demonstrated enhanced predictive power, driven by life quality score, week 2/4 response status, and week 4 dosage (accuracy = 78.02%, PPV = 81.48%, NPV = 72.97%). Logistic regression confirmed week 4 response status as a significant predictor of week 8 outcome (p < 0.005).

Week 4 emerges as a key decision point for escitalopram-treated MDD patients, where integration of baseline profiles, early response patterns, and dose parameters allows timely intervention. Our decision tree framework offers a methodological approach for dynamic decision points that warrant prospective validation and extension to other antidepressants.