The present paper examined the relationship between age at menarche and current practices of smoking, heavy drinking and drug use among 14-15 year old females in Canada. The prevalence of smokers, heavy drinkers and drug users in the study sample was approximately 22%, 38% and 26%, respectively. After adjusting of all potential confounders, no association was found between age at menarche and any of the substance use outcomes. School performance and relationship with the father, however, stood out as the main variables to be associated with smoking, heavy drinking and drug use.
The prevalence of substance use among 14-15 year old Canadian girls reported in this study is higher than previous Canadian estimates. Using the NLSCY data (cycle 3, 1998/1999), Hotton & Haans (2004) reported alcohol intoxication in the past 12 months among girls (12-15 years) was 19.0% as compared to 37.7% in the present study . In the same study, the prevalence of drug use in the past 12 months was half (18.3%) of the rate reported in this study. As for smoking, the results of the Canadian Tobacco Use Monitoring Survey (2003) suggest that 20% of the females aged 15-19 years were smokers .
Despite the high prevalence of substance use outcomes, age at menarche in the present study did not explain any variation in the current practices of smoking, heavy drinking and drug use. Previous literature, in contrast, has revealed that early pubertal timing increases the risk of substance use. A longitudinal study from the United States, revealed that early developing girls in the 7th grade were 2.7 times (95% CI: 1.6-4.4) more likely to have been drunk in the past 12 months, 2.5 times (95% CI: 1.5-4.3) more likely to have ever tried marijuana and 1.6 times (95% CI: 1.3-2.0) more likely to have ever tried cigarettes . A more comprehensive cross-sectional study by Patton et al. (2004) assessed the effect of pubertal development based on tanner stages (I to V) on substance use among 10-15 year old boys and girls in the United States and Australia . Based on the findings of the study, lifetime substance use was almost twice as high (OR = 1.7, 95% CI: 1.42-2.1) in adolescents undergoing mid-tanner stages (III) and three times (OR = 3.1, 95% CI: 2.4-4.2) as likely among those undergoing more advanced tanner stages (IV/V) as compared to their counterparts undergoing I/II tanner stages. Recent substance use, however, was less pronounced in adolescents going through stage III (OR = 1.4, 95% CI: 1.0-1.9) but more strongly associated with those in their stage IV and V (OR = 2.3, 95% CI: 1.7-3.3) .
The association between early puberty and substance use has been attributed to biological, psychological and social factors. It has been hypothesized that pubertal hormones have effects on the cognition and behaviour of the adolescents which consequently contribute to substance use [4, 13]. Early maturing girls might also engage in substance use because of the pressure of the misbalance between the adult appearance and the social and psychological immaturity [14, 15]. Girls who perceive their tempo of puberty as unusual from their peers might experience high levels of stress which might lead to inappropriate decisions, like substance use . Moreover, girls who mature earlier are thought to affiliate with older peers and hence imitate their behaviour [4, 5, 12].
Interestingly, in spite of the evidence of the previous literature and supporting hypotheses of the association between puberty and substance use, no relationship was found in the present study. The explanation for this finding is unclear without further quantitative studies to confirm these results. Qualitative studies on social and psychological changes experienced by early maturing Canadian adolescents are also warranted to identify other correlates or pathways to substance use in this higher risk population. Nevertheless, examining longitudinally the substance use trajectories among the different menarche groups adds further insight to the subsequent impact of age at menarche on adult substance use. A plausible explanation maybe that the emotional stress associated with early maturation might not have an effect on substance use. This is consistent with a Canadian study by Hotton and Haans (2004) that suggests that emotional problems do not influence adolescent substance use . In this study, the authors found that youth with emotional problems were less likely to be drunk in the past year than their counterparts . Further to the above, self-esteem which is associated with early puberty  was not found in the present study to be significantly associated with any of the substance use outcomes. Moreover, it has been hypothesized that as girls grow older the association between age at menarche and substance use practices dissipates . Since the present study assesses this association during mid-adolescence, the disagreement of the results with the literature might be due to the assessment of this association with substance use initiation at younger ages. Further to the above, this disagreement might be attributed to the multivariate analysis conducted in this study. As mentioned earlier, most studies examining the relationship between substance use outcomes and puberty were based on bivariate associations [1, 2, 4, 10, 12, 16, 17].
In this study, school performance and relationship with the father, rather than age at menarche, played an important role in explaining practices of substance use. The results are consistent with another Canadian study that was conducted using the NLSCY, cycle 3 data . Students having poor or very poor grades at school were 2.35 times (95% CI: 1.21-4.54) more likely to have drunk to intoxication than those having good or very good grades at school. Other studies have also shown an association between low school commitment and substance use [31, 32]. The Canadian study has also shown that hostile and negative parenting styles were associated with drug use (OR = 1.09, 95% CI: 1.02-2.16) and alcohol intoxication (OR = 1.11, 95% CI: 1.03, 1.19) . A review paper investigating the relationship between adolescent alcohol consumption and parenting strategies, revealed that good parent-child relationship delays alcohol initiation (p < 0.002) and decreases future levels of alcohol use (p < 0.001) .
The response rate at cycle 4 was 85.4%. However, the cross-sectional weights used in the analysis takes into account for the non-response. In the present study, girls for whom information on age at menarche and substance use outcomes was missing were excluded from the analysis. Therefore, the percentage of girls with complete data was 72.7%. Girls excluded from the analysis were compared in terms of socioeconomic, demographic, and child-related factors to those girls included in the analysis. No statistical difference, except for province, was detected between the two groups (data not shown). Girls with missing data were more likely to be residents of the Central provinces. Another limitation of the study is the cross-sectional design of the study where the temporal relationship between the outcomes and the covariates cannot be determined. The outcome variables, also, were measured in a limited time frame (within the past 12 months). Future studies examining longitudinally the substance use trajectories focusing on the frequency of substance use are warranted. Further to that, the substance use outcomes were dichotomous in nature. Since the three outcomes were assessed in the survey using different measures, a two-level variable allowed the comparability of the substance use outcomes. Some potential confounders were also missing from the analysis. Peer substance use, an important determinant of substance use, was collected in the NLSCY but not included in the analysis due to its high co-linearity with the outcome variables. Parental substance use was not available. Only PMK's smoking and alcohol drinking practices were collected in the NLSCY. Moreover, there was no access to academic records from schools and school performance had to be collected by the PMK. Finally, the study results are reflective of the years 2000/2001. Although more recent cycles of the NSCLY are available for the year 2006/2007, generalizability of such results would be limited to the original NLSCY cohort that was sampled in 1994. At each cycle, Statistics Canada creates cross-sectional weights to account for demographic changes in the population. These weights allow the generalizability of the results to the Canadian population. No cross-sectional weights, however, were created after cycle 4. Nevertheless, this is the first study that looks at the association between age at menarche and substance use in the Canadian population. The sample size of the study, the multivariate analysis and the decreased likelihood of recall bias stand out as the main strengths. Age at menarche was reported at an age (14-15 years) close to the menarche event. Based on Koo (1997), the accuracy of reporting age at menarche increases with a decrease in recall interval .