Cannabis lost its network hub role among Colombian students

BACKGROUND: The gateway hypothesis predicts that cannabis occupies a stable hub position linking legal and illicit substance use, yet rising cannabis prevalence alongside declining tobacco and alcohol use may be reshaping this co-occurrence structure. We examined whether that structure changed over eleven years. METHODS: Data were drawn from cross-sectional surveys of Colombian university students conducted in 2012, 2016, and 2023. For each wave we estimated a network of lifetime use of seven substances using mixed graphical models, in which each node is a binary lifetime-use indicator and each edge represents the conditional association between two substances after adjusting for all other substances, age, and socioeconomic stratum. We compared waves with the Network Comparison Test, a permutation test that evaluates two properties separately: whether the overall pattern of edge weights differs between two networks (network structure) and whether the sum of all absolute edge weights differs (global strength). We characterized each network with global and node-level centrality metrics. RESULTS: Network structure changed significantly between 2012 and 2023 and between 2016 and 2023, but not between 2012 and 2016; global strength fell from 12.15 to 8.98. All p-values are Holm-adjusted across comparisons. Cannabis centrality declined substantially: weighted degree fell from 6.19 to 2.02 and eigenvector centrality from 1.00 to 0.55. Tobacco replaced cannabis as the most central substance by 2023 despite declining prevalence. Non-prescription tranquilizer use tripled but remained peripheral. CONCLUSION: The co-occurrence structure of lifetime substance use among Colombian university students reorganized between 2016 and 2023, with cannabis losing its hub position. These patterns are relevant to Colombia's ongoing cannabis regulation debate, university prevention programming, and the growing challenge of non-prescription tranquilizer misuse.