[Comparison of two methods to identify factors associated with the onset of cannabis use in a cohort study]

To compare the utility of two multivariate analytic methods, case-control (CC) analysis and nested case-control (NCC) analysis in a cohort, to identify the factors associated with the onset of cannabis use.

A longitudinal cohort study of a sample of secondary school students (n = 1,056) in the first year of secondary school was carried out. Participating students were followed-up annually until the fourth year of secondary school. In the CC analysis, students in the fourth year who reported having consumed cannabis at some time were considered cases and those who had never consumed cannabis were considered controls. Logistic regression (LR) models were estimated. In the NCC analysis, cases in each year of follow-up were compared with a random sample of controls at risk in the same year and conditional LR models were estimated.

In the CC analysis, 6 variables in boys and 9 variables in girls in bivariate models and 3 variables in boys and 4variables in girls in multivariate models were identified. In the NCC analysis, 17 variables in univariate models and 4 in multivariate models were obtained in both boys and girls. The estimators of the CC analysis showed an average of 1.2-fold more variability.

A higher number of factors associated with cannabis use were identified in the NCC analysis and the estimators were more precise. NCC could be a more efficient option than CC analysis.