The current study examined the possible predictors of dropout during a five-week light treatment (LT) with a gradual advance protocol in 93 patients with unipolar non-seasonal depression and evening chronotypes by comparing their clinical characteristics and performing a logistic regression analysis. Nineteen out of ninety-three (20%) subjects (80% female, 46.5 ± 11.7 years old) dropped out during the 5-week light treatment. Treatment non-adherence (i.e., receiving LT for less than 80% of the prescribed duration) over the first treatment week predicted a five-fold increase in risk of dropout during light therapy (OR: 5.85, CI: 1.41–24.21) after controlling for potential confounders, including age, gender, treatment group, rise time at the baseline, patient expectation, and treatment-emergent adverse events. There is a need to incorporate strategies to enhance treatment adherence and retention in both research and clinical settings. Chinese clinical trial registry (ChiCTR-IOR-15006937).