Our study represents one of the first examples in which specific genes and environmental risks are jointly studied to understand how environmental risk might modify the genetic risks for nicotine dependence. Testing the hypothesis that level of parent monitoring modifies the known risk of nicotine dependence associated with two distinct variants in nicotinic receptors (CHRNA5 and CHRNA3), we found a statistically significant interaction between level of parent monitoring and SNP rs16969968 (CHRNA5) but not SNP rs3743078 (CHRNA3).
A family environment which is characterized by parents who sometimes, hardly ever or never monitored their middle school aged children’s whereabouts fell into the lowest quartile of parent monitoring whereas higher quartiles are characterized by parents who monitored their children most or all of the time. For example, 72.6% of subjects in lowest quartile parent monitoring reported sometimes, hardly ever or never left a note for their parents about where they were going, in contrast with only 5.1% of subjects in higher quartiles. About 42.3% of subjects in lowest quartile parent monitoring reported parents sometimes, hardly ever or never knew if they came home an hour late on weekends, in contrast with only 2.8% of subjects in higher quartiles. In addition, we found in our data that the association between levels of parent monitoring and nicotine dependence is not linear. Individuals with the lowest quartile of parent monitoring was associated with a significantly increased risk and there was no gradient in the remaining three quartiles. This is consistent with a prior study showing only the lowest quartile level of parent monitoring was associated with a sustained higher risk of drug initiation (24
). Furthermore, when we tested the interactions between the risk genotype and each of the quartiles of parent monitoring, we found a similar non-linear effect showing the interaction only between the risk genotype and the lowest quartile, without gradient in the remaining three quartiles.
The finding that the genetic risk associated with rs16969968 was reduced in subjects with high parent monitoring, and increased in subjects with low parent monitoring suggests that exposure to low parent monitoring during the middle school period creates an environment that allows for greater opportunity to express this genetic predisposition. These results are in line with findings from the Finnish twin studies, which suggest that parent monitoring modifies the importance of genetic influences on adolescent smoking; genetic influences assuming a far greater importance at low levels of parent monitoring (28
In contrast, however, the genetic risk associated with another SNP rs3743078 did not vary with level of parent monitoring. Protective genotypes of rs3743078 and parent monitoring showed independent and additive associations with nicotine dependence. Thus these results indicate the likely complexity of the interplay between specific genetic and environmental risks for nicotine dependence.
The results of this study should be interpreted in the context of several limitations. First, our assessment of parent monitoring was based on self report, as elicited with multi-item questions, which is less comprehensive than multi-informant assessments from both parents and children (37
). Second, the retrospective nature of this study’s ascertainment of childhood environmental risk exposures is subject to recall bias, which may have affected the validity and reliability of reported parent monitoring. However, this concern is reduced somewhat given that our main effects association of parent monitoring with nicotine dependence is consistent with longitudinal work among adolescents (24
). Third, our measure of parent monitoring is one of various instruments assessing parent monitoring, and does not differentiate different aspects of parenting practices such as parent involvement and parent discipline (24
). Fourth, we also have to be cautious in labeling parent monitoring solely as an environmental risk, as parent monitoring itself may be influenced by both genetic and environmental factors, reflecting both genetically-influenced characteristics of the parents (39
), though neither genetic variant examined here was correlated with parent monitoring (r=0.03, p=0.18 for rs16969968; r=−0.03, p=0.18 for rs3743078). Fifth, we found no significant difference in the associations between parent monitoring and nicotine dependence between the genders in this data by testing the interaction between gender and parent monitoring (p=0.071). The genetic risks associated with the risk genotype rs16969968 were not statistically different across gender (OR=2.15 for men; 1.82 for women). Because of the possible difference in parent monitoring across gender, we redefined our lowest quartile parent monitoring within each gender group to ensure that the comparisons are made within the same gender group. In gender-stratified analyses, we found the similar pattern of stronger gene-modifying effects of parent monitoring in females than in males. In women, OR for risk genotype = 1.49 with higher monitoring increased to 4.08 with low parent monitoring. In men, OR for risk genotype =2.03 with higher monitoring increased to 2.76 with low parent monitoring. However, there is a potential gender difference in the interaction between lowest quartile parent monitoring and risk genotype of rs16969968, which may be beyond the statistical power of the current study sample. However, there are many more environmental and personal risk factors (e.g., childhood adversity, peer use, parental smoking, and personality characteristics) as well as specific genes what influence the development of nicotine dependence than were examined here (43
). The variance (r2
) explained by these two SNPs and parent monitoring is 4.2%. We anticipate the discovery of many more genetic and environmental risks associated with the development of nicotine dependence. More complex and complete models will need to be developed in the future. Another limitation is the ascertainment bias. The estimation of gender-specific prevalence of nicotine dependence needs to take into account the ascertainment bias. However, we were able to control for the potential confounding effects of gender in our study when examining the association between genetic/environmental risks and nicotine dependence in multivariate regression models. Caution is needed in interpreting our findings on the interplay between genetic risk and parent monitoring because replications in independent samples are necessary.
The statistical power in testing gene-environmental interaction is a function of sample size, prevalence and relative risk of risk genotype, prevalence and relative risk of environmental factors, type I error, and hypothesized G*E interaction relative risk. The power of our study sample (N=2027) to detect a gene-environment interaction relative risk of 2.50 or higher between SNP rs16969968 (risk genotype frequency 12.5%) and low parent monitoring (prevalence 22.9%) is estimated to be 0.70 at α=0.05. We found a significant interaction between rs16969968 and parent monitoring. Our study sample has the sufficient power greater than 0.80 to detect a gene-environment interaction relative risk of 1.85 or higher between SNP rs3743078 (protective genotype frequency 40.2%) and low parent monitoring (prevalence 22.9%) at α=0.05. However, we found no significant interactions between rs3743078 and parent monitoring.
While acknowledging the limitations of our work, we note that this work complements prior research on parent monitoring (24
), and provides novel and significant findings. In addition to estimating how parent monitoring may alter the overall genetic risk, we extended the results to adults, and more importantly to the study of specific risk genotypes. It is also important to note that the difference in results between our two SNPs suggest that environmental modification of genetic risk is not uniform. Therefore, it is critical to look at specific genetic risks as some may be modified by an environmental factor while other genetic risks may not be.
Although replication of this study’s results are needed, our findings strengthen the case for development and rigorous testing of prevention strategies that target parent monitoring as a potentially malleable risk factor for nicotine dependence, in particular among individuals with the high risk genotype of rs16969968. It is possible that those with moderate to high parent monitoring during middle school ages have significantly more limited opportunities for experimenting with tobacco or smoking regularly, and the likelihood of developing nicotine dependence for those individuals is reduced despite the increased genetic risk. This possibility is consistent with previous studies that suggested that higher levels of parental involvement/reinforcement might help shield youths from substance exposure opportunity from mid-childhood to early adulthood, indicating a possible mechanism of protection against early-onset substance involvement (38
), and which creates an environment that minimizes the opportunity to express genetic predispositions. However, when adolescents receive low parent monitoring, it creates a different environment that allows for greater opportunity to express genetic predispositions to develop nicotine dependence. Though parent monitoring is good for all, in the context of the genetic risk for nicotine dependence, it is more salient for those with the rs16969968 risk genotype.
These results are beginning to fill gaps in our understanding of the joint association of specific genes and environmental risk factors with nicotine dependence. The presence of an interaction between identified risk genes and environmental factors suggests that environmental risk factors may accentuate the effect of specific risk genes for nicotine dependence. Identification of environmental factors that might modify the expression of specific risk genes will allow for the development of more effective interventions. These analyses lay the ground work for future prospective studies.