Accurate assessment of medication adherence is important for understanding pharmacotherapy outcomes across all phases of adolescent substance use disorder (SUD) clinical trials. The objective of this study was to describe and assess thepairwise concordance between three commonly used non-biological medication adherence assessment methods in adolescents who use alcohol to inform the selection of medication adherence measures for use in future youth SUD trials. Participants (N = 32, 17-19-years-old) took N-acetylcysteine and placebo, in a randomized cross-over design, for 10 days each. Medication adherence was assessed (20 days total) via pill count, medication videos submitted twice daily, and the Medication Event Monitoring System (MEMS®). Lin's Concordance Correlation Coefficient (CCC) assessed concordance and Bland-Altman plots are reported. Linear mixed-effects models with main effects of medication, treatment block (first medication, second medication), and sequence were also run. Medication videos yielded the lowest (64%) and pill count yielded the highest (89%) adherence estimates. CCC values indicated poor correspondence, except between pill count and MEMS. The Bland-Altman plots showed good pairwise agreement between all methods. Linear mixed-effects models indicated a difference between the first and second cross-over medication, with adherence estimates being lower for the second medication, regardless of whether it was N-acetylcysteine or placebo. The study yielded important and practical information. First, incorporating more than one method of adherence assessment may capture estimated floor and ceiling adherence in the absence of a biological marker. This is particularly relevant for remote or hybrid studies where bio-marker collection is challenging. Selection of the assessment methods will depend on study goals. Second, the continuation of medication adherence research can benefit each phase of clinical trials and inform rigorous pharmacotherapy evaluation.