Latent class analysis of 496 cannabis users in São Paulo identified four distinct user profiles, with about 30% in high-risk categories marked by early onset, frequent use, and polydrug patterns.
public health researchers, substance use epidemiologists, harm reduction program designers
What the researchers found
A four-class model identified distinct cannabis user profiles: a Polydrug class (26.2%) and a Former class (5.9%) with earlier onset and highest frequency, and two lighter-use groups. The Polydrug class had higher odds of other drug use (OR 3.0), tobacco (OR 2.5), heavy episodic drinking (OR 1.8), and alcohol use disorder (OR 1.5).
Why it matters
Not all cannabis users face the same risks. Identifying distinct user profiles helps target prevention and harm reduction efforts toward the subset of users who are most vulnerable to negative outcomes.
The numbers in context
5,037 total survey participants. 496 cannabis users analyzed. Four classes identified. Polydrug class (26.2%): OR 3.0 for other drugs, OR 2.5 tobacco, OR 1.8 heavy drinking, OR 1.5 AUD. Former class (5.9%): earliest onset and highest historical frequency. About 30% at increased risk.
How the study worked
Data from the São Paulo Megacity Mental Health Survey (n=5,037). Latent class analysis of 496 cannabis users based on age of onset, use frequency, tobacco consumption, heavy episodic drinking, alcohol use disorder, and substance use disorder. Logistic regression assessed correlates.
What this study cannot tell us
Cross-sectional design limits causal inference. Self-reported substance use. São Paulo sample may not represent all of Brazil. Latent class analysis findings are sample-dependent and may not replicate in other populations.
How to read the evidence
Population-based survey with sophisticated statistical methodology, but cross-sectional design and single-city sampling limit evidence to moderate.
When this study was published
Data from the São Paulo Megacity Mental Health Survey.
The bigger picture
This population-level classification study from Brazil adds to the global evidence that cannabis use is not monolithic. A minority of users account for a disproportionate share of harm, a pattern consistent with findings from other countries.
Questions still open
- Can the high-risk classes be identified early enough for prevention?
- Would these same classes emerge in different cultural contexts?
Common questions
What defines high-risk cannabis use?
Are most cannabis users at high risk?
Read the original research
Cannabis use patterns and different phenotypes in relation to other drugs use: latent class analyses from the Sao Paulo Megacity Mental Health Survey.
Revista brasileira de psiquiatria (Sao Paulo, Brazil : 1999), 47, e20243833
Citation
Silveira, Camila Magalhães; Castaldelli-Maia, João Maurício; Siu, Erica Rosanna; Viana, Maria Carmen; Wang, Yuan-Pang; Andrade, Laura Helena. (2025). Cannabis use patterns and different phenotypes in relation to other drugs use: latent class analyses from the Sao Paulo Megacity Mental Health Survey.. Revista brasileira de psiquiatria (Sao Paulo, Brazil : 1999), 47, e20243833. https://doi.org/10.47626/1516-4446-2024-3833