The declining centrality of cannabis in networks of lifetime substance use among Colombian University students, 2012-2023.

Cardozo F.; Scoppetta O.; García-Mejía N.; Baral A. · The International journal on drug policy · 2026

**Background & Methods** This longitudinal cross-sectional network analysis examined changes in lifetime substance use co-occurrence patterns among Colombian university students across three survey waves (2012, 2016, 2023). Researchers used mixed graphical models to construct networks of seven substances, with edges representing conditional associations between substances adjusted for age and socioeconomic factors, followed by Network Comparison Tests to evaluate structural changes between time periods. **Key Findings** - Cannabis centrality declined substantially over the 11-year period, with weighted degree decreasing from 6.19 to 2.02 and eigenvector centrality declining from 1.00 to 0.55, indicating loss of its hub position as a central linking substance between legal and illicit drug use. - Network structure and global strength changed significantly between 2016 and 2023 (but not between 2012 and 2016), with global strength falling from 12.15 to 8.98, suggesting reorganization of substance co-occurrence patterns occurred primarily in the most recent period. - Tobacco emerged as the most central substance by 2023 despite declining overall prevalence, while non-prescription tranquilizer use tripled but remained peripherally positioned within the network. **Dosage & Administration** Not reported. **Safety & Adverse Effects** Not reported. **Evidence Quality** This study provides moderate-quality evidence with several limitations. The analysis is restricted to university students, limiting generalizability to the broader Colombian population or other demographics. Cross-sectional survey methodology cannot establish causation or temporal relationships between individual substance use transitions. The study relies on self-reported lifetime use data, which may be subject to recall bias or social desirability bias. Strengths include longitudinal comparison across 11 years, rigorous statistical methods (network analysis with permutation testing), and adjustment for relevant covariates. The Network Comparison Test provides valid inference for structural differences. Findings are observational and contextual to Colombia's specific substance availability, regulatory environment, and university population characteristics, potentially limiting international applicability.

More recent cannabis research