Background Confounding introduced by gene–environment correlation (rGE) may prevent one from observing a true gene–environment interaction (G × E) effect on psychopathology. The present study investigated the interacting effect of the BDNF Val66Met polymorphism and stressful life events (SLEs) on adolescent depression while controlling for the rGE by two means: separating pure environmental factors (independent SLEs) from the environmental factors under partial genetic control (dependent SLEs) and adopting a prospective longitudinal design. Methods A total of 780 pairs of Chinese twins, aged 11–17 years (mean = 13.6, SD = 1.8) at intake, were followed up twice. Self‐reported depression symptoms at Time 1 and Time 2 were assessed by the Children's Depression Inventory (CDI). SLEs occurring between Time 1 and Time 2 were assessed by a self‐reported checklist. SLEs were differentiated into independent and dependent ones and were validated by heritability analyses using twin design. The interacting effects between the BDNF Val66Met polymorphism and numbers of SLEs (total SLEs and independent SLEs) on intraindividual change of depression symptoms were examined. Results After controlling for sex, age, age square, and Time 1 depression, both total SLEs × BDNF Val66Met genotype and independent SLEs × BDNF Val66Met genotype significantly predicted Time 2 depression. Val allele carriers (Val/Val and Val/Met) were more susceptible to the detrimental effects of stress. Conclusions There is a true G × E effect underlying the observed interaction between BDNF Val66Met polymorphism and environmental stress on depression.