AIMS - To synthesize the available evidence on predictors of adolescent tobacco use.
DESIGN - Meta-analysis was conducted on the empirical findings of published and unpublished studies of the natural development of tobacco use that used prospective multi-wave panel designs.
PARTICIPANTS - The research literature that was analyzed included 106 reports on 64 studies representing data from a total of 145,750 study subjects; 1261 prospective and cross-sectional effect sizes were computed from these studies and used in the meta-analysis.
MEASUREMENTS - Product-moment correlations were analyzed examining the strength of the relationships between predictor variables and current and later tobacco use. In addition, findings reported as 2 x 2 contingency tables were analyzed to examine conditional relations and estimate the positive predicted values (PPV) and sensitivity indices for the predictive relationships.
FINDINGS - The mean correlations for 17 different categories of predictors and current or later tobacco use ranged from -0.08 for race to 0.52 for prior tobacco use and were significant and positive except for race. Analysis of the conditional relationships showed that PPV for tobacco use ranged from a mean of 0.18 for predictors related to personal skills and knowledge (i.e. 82% of those 'at risk' on this construct did not use tobacco) to 0.70 for use of tobacco or other substances by parents.
CONCLUSIONS - Current use of tobacco and other substances by youths, and use among their peers, showed stronger relationships with later tobacco use than other examined predictors. Combined with other predictive risk factors, these relationships are sufficiently strong to be useful in identifying for intervention those children most likely to become habitual tobacco users.