A popular longitudinal mediation model is the latent growth curve mediation model (LGCMM). With the LGCMM, researchers can examine how treatment longitudinally influences the level (i.e., intercept) and change (i.e., slope) of the outcome through influencing the level and change of the mediator. Despite the popularity, causal mediation analysis with the LGCMM has been understudied. To reduce the research gap, in this dissertation, I studied causal mediation analysis with the bivariate LGCMM where the treatment is time-invariant whereas the mediator and outcome are both time-varying. To handle the post-treatment confounding issue posed by the dependence between mediator intercept and slope, I extended the interventional effect definition for causal interpretation of the indirect effects via the mediator intercept (or slope) alone. Furthermore, I defined and identified an interventional indirect effect not attributable to either the mediator intercept or slope alone but due to their mutual dependence. For effect estimation, I proposed the interaction LGCMM, which incorporates interactions among treatment, mediator intercept, and mediator slope and allows the residual covariance of mediator intercept and slope to differ between treatment vs. control groups. A Bayesian approach was proposed to estimate the model. Simulation results showed that the Bayesian approach yielded satisfactory estimates and inferences of the interventional direct and indirect effects in the interaction LGCMM with N ≥ 400 when T ≥ 6 or with N ≥ 200 when T ≥ 12. The results also demonstrated that when there are true interaction effects between the mediator intercept and slope, the traditional LGCMM ignoring such interactions can produce biased estimates and inaccurate inferences of the interventional direct and indirect effects. The developed method was applied to data from a real-life longitudinal study for illustration. The current study provides insights on the causal mediation effects in LGCMMs and adds to researchers' toolbox of investigating longitudinal pathways.