Issues of validity in longitudinal studies of youth tobacco use

Abstract

In the Stockholm region, a prospective longitudinal study: BROMS Cohort Study (Children s Smoking and Environment in the Stockholm County) was conducted between 1997 and 2005, with the primary aim to investigate determinants of adolescents´ tobacco use. It was based on a sample recruited in schools with junior high grades. This thesis highlights issues of validity in cohort studies based on samples of children and adolescents. Research questions of interest were: determinants of participation at the school, family and individual levels; validity of reports of smoking and snus use among adolescents; methods for assessment of symptoms of nicotine dependence among adolescent tobacco users; stability of retrospective recall of smoking during pregnancy among the adolescents mothers. All papers in this thesis are based on data from the BROMS Study.

At the school level, the socio-economic characteristics of the area, the size of the schools, but not the presence of a tobacco policy predicted school participation. At the parental level, education and average social status in the residential area was associated with parental consent. At the level of the adolescents, the attrition over 7 years follow-up was low as only 6% of the subjects dropped out of the study permanently, and the proportion participating in all follow-up surveys was close to 70%. Many psychosocial traits and behavioural factors beside tobacco use predicted lack of sustained participation. Significant predictors were: male gender, an increasing number of recent stressful events, most of them connected with unstable family conditions, school truancy and low school performance. In a sub-sample of 520 adolescents at aged 15 years, self-reports of cigarette and smokeless tobacco (snus) use were compared with saliva cotinine concentrations (cut-point of 5 ng/ml). Overall, the concordance between self-reports and saliva cotinine levels was high. Disagreement was found in subjects reporting occasional, infrequent or irregular tobacco use.

At the age of 17 years, lifetime experience of nicotine dependence and withdrawal symptoms were reported by 466 exclusive smokers, 209 exclusive snus users and 144 dual users. A significantly higher proportion of dual users and exclusive snus users reported symptoms of nicotine dependence (withdrawal excluded) compared to exclusive cigarette smokers. Dual users reported the highest frequency of most withdrawal symptoms. Craving for tobacco was the most frequently reported withdrawal symptom among all types of users. Overall high stability and consistency of retrospective recall of smoking during pregnancy by the adolescents mothers was found when this was compared with reports given by the same mothers at the time of their pregnancies as recorded in the Swedish Medical Birth Registry. Discrepancies were more frequent among women with low-level pregnancy smoking and among women who had changed their smoking behaviour over time.

In summary, participation of schools and families in longitudinal studies involving youths cannot be predicted in a straightforward way as determinants of participation differ at the organisational and individual level. Hence, consideration on selection bias and generalisation must take into account study-specific hypotheses. Retrospective recall of smoking in pregnancy is reliable, but instability of recall is likely to occur for low-level or irregular smoking. Selfreports of current tobacco use in mid-adolescence are reliable, when elicited through questions that take into account the specific patterns of tobacco use in adolescence, as was the case in this cohort study. Smokeless tobacco use in adolescence is associated with symptoms of nicotine dependence at least as frequently as cigarette smoking is. Young tobacco users who combine smokeless tobacco with smoking obtain particularly high scores on scale items assessing nicotine dependence. In longitudinal studies of youths, baseline psycho-social characteristics of the study subjects can be used to predict, thence to prevent or minimize losses to follow-up.