The Effect of Maternal Smoking on Offspring Smoking Is Unrelated to Heritable Personality Traits or Initial Subjective Experiences

Abstract

Introduction

Maternal smoking is a risk factor for offspring smoking. Lifetime maternal smoking vs. prenatal tobacco exposure (PTE) appears to act through different mechanisms. This study tested the hypothesis that maternal smoking measures’ effects on offspring smoking could be attributable to hereditary mechanisms: personality traits (novelty-seeking, impulsivity, neuroticism, and self-esteem) and initial subjective smoking experiences (pleasurable, unpleasurable, and dizziness).

Methods

Data were drawn from the Social and Emotional Contexts of Adolescent Smoking Patterns study, an 8-year longitudinal study of 9th or 10th graders at baseline (≈age 15) who experiment with smoking (<100 lifetime cigarettes; N = 594) at baseline. The young adult smoking frequency at the 8-year follow-up (≈age 23) was examined as a function of baseline characteristics (heritable trait, maternal smoking, PTE, and sex) and baseline smoking frequency and nicotine dependence. Structural equation models determined whether the inclusion of each heritable trait among offspring confounded the effects of maternal smoking (PTE or maternal smoking) on offspring smoking and nicotine dependence.

Results

Impulsiveness was associated with intermediate adolescent smoking frequency (B = 0.135, SD = 0.043, p = .002) and nicotine dependence (B = 0.012, SD = 0.003, p < .001). Unpleasurable first experience (B = 0.886, SD = 0.374, p = .018) and dizziness (B = 0.629, SD = 0.293, p = .032) showed a trend with intermediate smoking frequency that was nonsignificant after correcting for multiple comparisons. These traits did not confound maternal smoking’s effects.

Conclusions

None of the heritable traits examined in this model explained the effect of maternal smoking measures on adolescence or young adulthood offspring smoking. Further research is needed to elucidate the mechanism by which PTE and maternal smoking are linked to offspring smoking.

Implications

Prenatal tobacco exposure (PTE) and mother’s lifetime smoking present separate and independent risks for offspring smoking; however, their mechanisms seem unrelated to heritable personality traits and initial subjective smoking experiences. These findings have implications for separate screening strategies tailored to different age groups, especially related to PTE’s risk of smoking in young adulthood. Additionally, these findings add to the known risks of maternal smoking. Further research is needed to understand the mechanism underlying the risk posed by maternal lifetime smoking and PTE on offspring smoking behavior.

Introduction

Cigarette smoking remains the primary cause of preventable death in the US.¹ Although smoking rates among adolescents have declined in recent years,² cigarettes have become deadlier, particularly with respect to lung cancer,¹ underscoring the continued public health harms of cigarettes. Since the majority of individuals who smoke initiate during adolescence, prevention efforts require understanding the risk factors and mechanisms for smoking among adolescents.³

Maternal smoking is a well-known risk factor for smoking. Adolescents with mothers who smoke are more likely to smoke themselves,⁴ regardless of when the mother smoked.⁵ That is, the risk imposed by a mother ever smoking is not increased by continuing to smoke into the offspring’s adolescence.⁵

In addition to maternal lifetime smoking prenatal tobacco exposure (PTE)⁵ is also relevant. The effect of PTE on youth smoking and nicotine dependence outcomes is well-known,⁵ but there is considerable debate over the particular environmental,^6,7 hereditary,^8,9 or epigenetic¹⁰ mechanisms that may explain its link with offspring smoking. Additionally, the evidence is mixed about whether the risk posed by PTE is additional to the risk conveyed by confounding family- and environmental-level factors¹¹: some studies find that PTE does not appear to act through intrauterine mechanisms,¹² while others find effects on offspring over and above correlated family level effects when exposure measurement is of high quality.¹³ While there is considerable evidence that some of the relationship between PTE and youth smoking is hereditary, little is known about the specific hereditary mechanism involved.

Previous research from our group has utilized structural equation models (SEMs) to understand the mechanisms of maternal smoking measures (which we use here as an umbrella term to indicate maternal lifetime smoking, or PTE, or both) in adolescent and young adult smoking behavior and nicotine dependence.^5,14,15 Specifically, PTE is directly associated with smoking in young adulthood, but not adolescence; and maternal lifetime smoking has an effect that is mediated through adolescent smoking behavior and nicotine dependence. Thus, PTE and maternal lifetime smoking act through distinct mechanisms with different timing effects. PTE’s effect, specifically during young adulthood, possibly suggests a teratologic, genetic, or hereditary mechanism,⁵ as environmental mechanisms (e.g. socioeconomic status, behavioral modeling, access to cigarettes) would be expected to appear first during adolescence rather than be dormant until young adulthood; on the other hand, genetic effects can become stronger in young adulthood.¹⁶ However, genetic variants associated with nicotine dependence (nicotinic acetylcholine receptor [CHRN] single nucleotide polymorphisms) do not explain this relationship,¹⁴ leaving it unclear which mechanism (hypothesized to be hereditary) PTE acts through. The current paper expands on this previous research to explore the contribution of several heritable traits to the relationship between PTE and youth smoking using the same base SEM model.^5,14,15

One possible type of mechanism for the association between maternal smoking measures and offspring’s smoking is through heritable personality traits, sometimes conceptualized as a predisposition to problem behavior,¹⁷ that separately influence both the mother’s smoking during pregnancy and offspring smoking behavior. Specifically, heritability has been reported for novelty-seeking temperament (60% heritable),¹⁸ impulsivity (35–62%),¹⁹ neuroticism (15%),²⁰ and low self-esteem (40–62%),²¹ which are all risk factors for smoking.²² Initial subjective smoking experiences (both pleasurable and unpleasurable experiences, as well as dizziness), are also heritable (35%, 31%, and 37%, respectively)²³ and are associated with later smoking behavior.^23,24

Of particular interest in answering these research questions is the subpopulation of adolescents who smoke experimentally, as it is important to ultimately provide earlier intervention efforts to prevent the development of established smoking. One salient factor is early-emerging nicotine dependence, which can occur in some adolescents soon after smoking initiation and well before daily smoking habits are established.²⁵ Since this early emerging nicotine dependence predicts subsequent smoking behavior over and above prior smoking behavior,²⁵ our previous work has focused on a cohort of adolescent experimenters to understand risk factors and mechanisms for established smoking.

This paper extends our previous research on maternal and adolescent smoking by examining heritable traits as potential mechanisms that explain the previously observed direct relationship between maternal smoking measures and young adult smoking behavior^5,15 utilizing a longitudinal cohort of adolescent experimenters followed into young adulthood. The study aims to examine (1) whether heritable traits are associated with adolescent and young adult smoking, and if so, (2) whether this association mediates the respective effects of maternal smoking measures or is statistically independent of them. We do not include father’s smoking, as it is not significantly associated with offspring smoking in this sample.¹⁵

Material and Methods

Participants

Data were drawn from the Social and Emotional Contexts of Adolescent Smoking Patterns (SECASP) Study, which has been described previously.¹⁵ The SECASP study samples a cohort of N = 1263 adolescents who smoked experimentally (defined below), and a random sample of nonsmoking adolescents, in 9th or 10th grade at baseline. This cohort was followed longitudinally over 8 years, approximately annually. At each wave, participants completed a traditional questionnaire on smoking behavior, nicotine dependence, and other personal characteristics. In addition, a parent completed a questionnaire at baseline from which PTE was obtained. Baseline surveys were used for demographic characteristics.

Adolescents who, at baseline, had smoked under 100 cigarettes in their lifetime (N = 594; low-exposure group) were included in the current analyses (“experimental smoking”), consistent with the preceding studies’ emphasis on understanding the early stages of smoking addiction.^5,14

Measures

Mother’s lifetime smoking status was proxy-reported by adolescents at baseline during the household inventory questions, using a checkbox with three response options (“She smokes,” “She is an ex-smoker,” and “She has never smoked”), and was dichotomized into lifetime smoking (currently or formerly smoking vs. never-smoking adolescents), based on previous studies.^5,14,15

PTE was assessed via the parent questionnaire per trimester, in a manner that maximizes the accuracy of retrospective reports.¹³ In this study, PTE is defined as “persistent” PTE, which is any smoking during at least two trimesters. This definition is based on previous research showing no effect of smoking only one trimester⁵; this is consistent with the majority of women quitting upon discovering their pregnancy, with corresponding protective effects on fetal neurodevelopment.²⁶ PTE is a subset of maternal lifetime smoking, and thus the effects of PTE can only logically be estimated among offspring of mothers with positive lifetime smoking status.

Adolescent smoking frequency was assessed as the number of days in the past 30 on which the participant smoked. Young adult smoking frequency (at the 8-year follow-up wave) was used as the outcome, and baseline smoking frequency is used as a predictor. In addition, an aggregate “intermediate” value to capture smoking frequency across adolescence was created as an average of all the intermediate time points for which the adolescent was under 18 years old.

Adolescent nicotine dependence was assessed with the Nicotine Dependence Syndrome Scale (NDSS), using a modified version that has been validated for use in adolescents.²⁷ Each item was assessed on a scale of 1 (“not at all true”) to 4 (“very true”) and a total score was derived by averaging the 10 items. Nicotine dependence at baseline was used as a predictor, and an aggregate intermediate value across adolescence was calculated as described above.

Biological sex was self-reported at baseline with possible response options of male (0) and female (1).

Heritable Traits

Self-Esteem

Baseline self-esteem was assessed with five items from the Rosenberg Self-Esteem Scale (RSE; Cronbach’s alpha = 0.84).²⁸ The full RSE is a 10-item rating scale that is widely used with adolescents to assess global self-regard. The five items used here were selected based on unpublished psychometric analysis and included the items on feeling like a failure, having a positive attitude towards oneself, feeling useless, thinking oneself is no good, and being satisfied with oneself. Items were scored on a four-point scale from 1 (“strongly disagree”) to 4 (“strongly agree”), with appropriate reverse-coding (items 1, 3, and 4 above); responses to the items were averaged to create a total scale score, with higher scores indicating higher self-esteem.

Impulsivity

Baseline impulsivity was assessed using a modified version of the Barratt Impulsiveness Scale (BIS-11; Cronbach’s alpha = 0.81).²⁹ Eleven items are scored on a four-point scale from 1 (“rarely”/“never”) to 4 (“almost always”/”always”). The total BIS score was calculated by summing three categories of impulsiveness (attentional, motor, and nonplanning), where higher scores indicate higher impulsivity.

Neuroticism

Baseline neuroticism was assessed using the Eyesenck Personality Inventory (Cronbach’s alpha = 0.79).³⁰ Adolescents are asked to respond dichotomously (1 = “yes”, 0 = “no”) to each of 20 items reflecting unstable, moody, and anxious personality characteristics. Higher values on the summed scale indicate higher levels of neuroticism or anxiousness.

Novelty Seeking

Baseline novelty-seeking temperament was assessed using the Tridimensional Personality Questionnaire (TPQ; Cronbach’s alpha = 0.73), modified for use in adolescents.³¹ The scale addresses personality characteristics such as seeking thrills, excitement, preferring to act on feelings of the moment, without regard for rules and regulations. Eight items were collected on a scale of 1 (“not at all true”) to 4 (“pretty true”) and averaged into a final score where higher scores reflect higher levels of novelty seeking.

Initial Subjective Smoking Experiences

Baseline subjective experiences with their first cigarette were assessed using a seven-item scale developed by Pomerleau et al.³² Pleasurable (Cronbach’s alpha = 0.79) and unpleasurable (Cronbach’s alpha = 0.75) subjective experiences were assessed as separate subscales. Each individual item was assessed on a scale of 1 (“none/not at all”) to 4 (“intense/very much”), and items for each subscale were averaged.

Dizziness

Dizziness (“during the first time you tried smoking, did you feel dizzy?”) was assessed on a four-point scale, from 1 (none) to 4 (intense/very much). Dizziness is one item in the scale for initial subjective smoking experiences. Dizziness was examined as a separate variable in this study, as factor analysis has shown this to emerge as its unique factor, distinct from other factors of initial subjective smoking experiences.²³ Additionally, dizziness alone is associated with continued smoking and nicotine dependence²⁴ and is partly heritable.²³

Analyses

The current model builds on previously published SEMs^5,14,15 that were updated to include heritable traits as an additional baseline predictor (base model shown in Figure 1), with each trait examined in a separate model.^5,14 The outcome of smoking frequency in young adulthood was examined as a function of baseline characteristics (heritable trait, mother’s lifetime smoking status, PTE, and sex) and baseline smoking frequency and nicotine dependence. For each predictor, the model included pathways to the ultimate outcome of smoking frequency in young adulthood, as well as pathways to adolescent smoking frequency and/or nicotine dependence.

Figure 1.

Base structural equation model. ^aHeritable personality traits (self-esteem, neuroticism, novelty-seeking, and impulsiveness) and heritable subjective first-time experiences (pleasurable, unpleasurable, and dizziness) were added one at a time in the base model to test for confounding of the relationship between maternal smoking (PTE or mother’s lifetime smoking) and offspring smoking, after controlling for other baseline variables.

Multiple final models were made by adding each heritable trait and subjective initial experience variable to the base model to assess confounding by each trait (see next paragraph). Model fit was assessed using several fit indices: chi square, Comparative Fit Index (CFI) ≥ 0.95,³³ Tucker-Lewis Index (TLI) ≥ 0.90,³⁴ and the root mean square error of approximation (RMSEA) ≤ 0.05.³⁵

Results were analyzed first with respect to whether each heritable trait had a significant main effect on smoking and nicotine dependence outcomes. The potential role of each significant pathway in explaining the effects of maternal smoking measures was then examined. First, we examined whether each significant pathway emerging from a heritable trait (pathway a→c in Figure 2) altered or eliminated (confounded) the effect of the maternal smoking measure on offspring smoking/nicotine dependence (pathway b→c in Figure 2). The two maternal smoking pathways of interest were: (1) PTE to young adult smoking frequency; and (2) maternal lifetime smoking to adolescent smoking frequency, due to its subsequent association with young adult smoking frequency. Specifically, unstandardized coefficients from the maternal smoking measures’ pathways (b→c in Figure 2) were compared between the base model and the model for each heritable trait (ie, presence vs. absence of pathway a→c). Partial confounding was defined as a change of ≥10%, indicating that the heritable trait explains some of the association between maternal smoking measures and offspring smoking.³⁶ Complete confounding is present when pathway b→c becomes nonsignificant after inclusion of pathway a→c (ie, a coefficient that does not significantly differ from zero), indicating that the trait completely explains the previously observed relationship between maternal and offspring smoking. It is possible that the pathway coefficient has an effect size change of greater than 10% but is not significant; this represents sample size limitations and therefore both the effect size and the significance were interpreted. Alternatively, if both a→c and b→c pathways are both significant, and the latter has changed by less than 10%, this indicates separate and statistically independent effects on young adulthood smoking. Finally, if the heritable trait pathway a→c is not significant, that indicates that the heritable trait has no additional effect on young adulthood smoking over and above the other terms in the model.

Figure 2.

Mediating pathway substructure. ^aHeritable traits included heritable personality traits (self-esteem, neuroticism, novelty-seeking, and impulsiveness) and heritable subjective first-time experiences (pleasurable, unpleasurable, and dizziness).^bMaternal Smoking Measure included either PTE or mother’s lifetime smoking status. ^cSmoking frequency included either adolescent smoking frequency or young adulthood smoking frequency.

Among models which show a significant pathway between the heritable trait and offspring smoking, the trait was then examined as a possible mediator of the maternal smoking measure’s effect on offspring smoking. Put another way, mediation occurs when the indirect effect (the product of predictor→mediator and mediator→outcome pathways) is nonzero; for this analysis examining trait as a potential mediator, these are the pathways between maternal smoking measure and trait, and between the trait and offspring smoking. Thus, the models above showing a significant pathway between trait and outcome meet one requirement for mediation, the other being that the heritable trait is associated with the maternal smoking measure. Thus, in cases with a significant heritable trait pathway, the remaining criterion for mediation was evaluated by examining the pathway between the maternal smoking measure and the trait.

All analyses were completed using R version 3.5.0, and SEM models were run using the “lavaan” package. The Bonferroni correction was used to correct for multiple comparisons across the seven different heritable traits, giving α = 0.007.

Results

Demographic Characteristics

Characteristics of the adolescent experimenters at baseline are shown in Table 1. Over half of the sample were females (57.9%) and the mean age at baseline was 15.6 years. Among this sample, 54.9% self-identified as non-Hispanic white, 19.7% as Hispanic, 16.3% as non-Hispanic black, and 9.1% as other or unknown. At baseline, the median days smoked was 1 day (interquartile range [IQR] = 0–4.5) in the past 30 days. Nearly half (272, 45.8%) of mothers reported current or former smoking and 72 (12.1%) reported any smoking during 2 (n = 2) or 3 trimesters (n = 70) trimesters. At young adulthood, the median number of smoking days was 0 (IQR = 0.0–17.0).

Table 1.

Baseline Demographic Characteristics of Experimenters^a

Baseline characteristics	Experimenters, N = 594	Missing values, n (%)
Female sex, n (%)	344 (57.9%)	0
Age, years, mean (range)	15.6 (13.9–17.2)	0
Race/ethnicity*  Non-Hispanic white, n (%)	326 (54.9%)	0
Hispanic, n (%)	117 (19.7%)
Non-Hispanic black, n (%)	97 (16.3%)
Other, or unknown, n (%)	54 (9.1%)
Baseline nicotine dependence, median (IQR)b	1 (0.6)	5 (0.8%)
Maternal lifetime smoking, n (%)c	272 (45.8%)	31 (5.2%)
Prenatal Tobacco Exposure, n (%)d	72 (12.1%)	148 (24.9%)
Baseline smoking frequency, median (IQR)e	1 (0–4.5)	1 (0.2%)
Adolescence, 4-year follow-up
Adolescent smoking frequency (intermediate aggregate),f median (IQR)e	3.6 (0.3–10.9)	17 (2.9%)
Adolescent nicotine dependence (intermediate aggregate),f median (IQR)b	1.2 (0.9–1.75)	18 (3.0%)
Daily smoking in adolescence, n(%)	4 (0.7%)	17 (2.9%)
Adolescent no smoking, n (%)	118 (20.55%)	17 (2.9%)
Young adulthood, 8-year follow up
Young adulthood smoking frequency, median (IQR)e	0 (0–17)	134 (22.6%)
Daily smoking in young adulthood n (%)	1 (0.2%)	134 (22.6%)
Young adulthood no smoking, n (%)	257 (55.9%)	134 (22.6%)
Heritable personality traits
Rosenberg self-esteem scale, median (IQR)g	3.0 (2.6–3.6)	0
Neuroticism, mean (SD)h	10.5 (4.2)	1 (0.2%)
Novelty seeking, mean (SD)i	3.6 (6.5)	2 (0.2%)
Barrett impulsiveness score, mean (SD)j	34.1 (7.4)	238 (18.8%)
Heritable subjective first-time experience
Pleasurable, median (IQR)k	1.7 (1.0–2.3)	16 (2.7%)
Unpleasurable, median (IQR)k	1.8 (1.5–2.5)	16 (2.7%)
Dizziness, median (IQR)k	1 (1.0–2.0	16 (2.7%)

IQR = interquartile range.

* Race/ethnicity was assessed on a six-point scale (non-Hispanic white; non-Hispanic Black, Asian or Pacific Islander, Hispanic, American Indian or Alaska Native, and Other or Unknown) and is presented here with “Other” containing Asian or Pacific Islander, American Indian or Alaska Native, and Other or Unknown race, due to low sample size.

^aExperimenters are defined as adolescents who, at baseline, have smoked fewer than 100 lifetime cigarettes.

^bNicotine dependence was assessed with a modified Nicotine Dependence Syndrome Scale (NDSS) on a scale of 1 (“not at all true”) to 4 (“very true”).

^cMaternal lifetime smoking was dichotomized as lifetime smoking (current or former smoking) vs. never-smoking.

^dPrenatal Tobacco Exposure (PTE) was defined as any smoking during two or three trimesters.

^eSmoking frequency was assessed as the number of days smoked in the last 30 days

^fIntermediate aggregate measures are derived from all available timepoints after baseline and until age 18 (see text).

^gAdolescent self-esteem was assessed with five items from the Rosenberg Self-Esteem Scale (RSE). Items are scored on a four-point scale from 1 (“strongly disagree”) to 4 (“strongly agree”); responses to the items were averaged to create a total scale score such that higher scores indicate higher self-esteem.

^hNeuroticism was assessed using the Eyesenck Personality Inventory, by summing 20 dichotomous items; higher values indicate higher levels of neuroticism or anxiousness.

ⁱNovelty seeking temperament was assessed using the Tridimensional Personality Questionnaire, these eight items were each collected on a scale of 1 (“not at all true”) to 4 (“pretty true”) and averaged into a final score.

^jImpulsivity was assessed using a modified version of the Barratt Impulsiveness Scale (BIS-11) Items are scored on a four-point scale where 1 is rarely/never and four is almost always/always. The total BIS score was calculated by summing three categories of impulsiveness (attentional, motor, and nonplanning), where higher scores indicate higher impulsivity.

^kHeritable subjective first-time experiences (pleasurable, unpleasurable, and dizziness) were assessed as separate subscales. Each individual item was assessed on a scale of 1 (“none/not at all”) to 4 (“intense/very much”), and items for each subscale were averaged.

Structural Equation Modeling

PTE’s effect was directly associated (ie, not mediated) with smoking frequency in young adulthood (B = 6.04, SD = 1.71, p < .001), while the relationship between a mother’s lifetime smoking and smoking frequency in young adulthood was fully mediated through intermediate smoking frequency (B = 1.95, SD = 0.69, p = .005) and nicotine dependence in adolescence (B = 0.18, SD = 0.052, p = .001).

Similar to findings from the baseline model, in all seven heritable trait models, PTE was directly associated with smoking frequency in young adulthood, and was statistically similar in magnitude to the base model (Table 2). Thus, none of these traits confounded PTE’s effect on smoking frequency in young adulthood. Similarly, none of the heritable traits were found to confound the effect of mother’s lifetime smoking on intermediate smoking frequency or nicotine dependence in adolescence. This indicates that none of these heritable traits explain the effect of PTE or maternal lifetime smoking on smoking behavior in their offspring.

Table 2.

Path Analysis Results for Heritable Personality Traits

Outcome variables (where path ends)	Base model	Self-esteem	Neuroticism	Novelty-seeking	Impulsiveness	Pleasurable	Unpleasurable	Dizziness
Fit statistics	χ2 = 0.00, CFI = 0.99, TLI = 0.96, RMSEA = 0.06	χ2 = 0.00, CFI = 0.99, TLI = 0.96, RMSEA = 0.05	χ2 = 0.00, CFI = 0.99, TLI = 0.96, RMSEA = 0.06	χ2 = 0.01, CFI = 0.99, TLI = 0.97, RMSEA = 0.05	χ2 = 0.00, CFI = 0.99, TLI = 0.96, RMSEA = 0.05	χ2 = 0.00, CFI = 0.99, TLI = 0.97, RMSEA = 0.05	χ2 = 0.01, CFI = 0.99, TLI = 0.97, RMSEA = 0.05	χ2 = 0.00, CFI = 0.99, TLI = 0.96, RMSEA = 0.05
Outcome: smoking frequency young adulthood	B (SE) p-value	B (SE) p-value	B (SE) p-value	B (SE) p-value	B (SE) p-value	B (SE) p-value	B (SE) p-value	B (SE) p-value
Heritable traita		0.22 (0.79), p = .783	−0.13 (0.12), p = .285	−1.08 (0.73), p = .139	0.11 (0.07), p = .114	1.03 (0.710), p = .145)	0.77 (0.63), p = .217	0.91 (0.50), p = .070
PTE	6.04 (1.71), p < .001	6.05 (1.71), p < .001	6.09 (1.70), p < .001	5.99 (1.70), p < .001	5.82 (1.71), p < .001	6.05 (1.70), p < .001	5.99 (1.70), p < .001	5.881 (1.71), p = .001
Mother smoking status	0.32 (1.15), p = .778	0.32 (1.15), p = .781	0.34 (1.14), p = .764	0.37 (1.14), p = .747	0.39 (1.14), p = .736	0.30 (1.14), p = .793	0.32 (1.14), p = .780	0.30 (1.14), p = .794
Sex	1.91 (1.02), p = .062	1.83 (1.06), p = .083	1.54 (1.07), p = .151	1.83 (1.02), p = .072	1.84 (1.02), p = .070	1.87 (1.02), p = .066	1.76 (1.02), p = .086	1.67 (1.02), p = .102
Baseline smoking frequency	−0.02 (0.13), p = .857	−0.03 (0.13), p = .847	−0.04 (0.13), p = .778	−0.02 (0.13), p = .850	−0.01 (0.13), p = .964	−0.04 (0.13), p = .763	−0.02 (0.13), p = .865	−0.04 (0.13), p = .770
Baseline nicotine dependence	1.71 (1.41), p = .227	1.73 (1.42), p = .223	1.91 (1.42), p = .180	1.75 (1.41), p = .215	1.58 (1.41), p = .262	1.27 (1.44), p = .381	1.67 (1.41), p = .237	1.55 (1.41), p = .272
Adolescent smoking frequency	0.52 (0.13), p < .001	0.52 (0.13), p < .001	0.54 (0.13), p < .001	0.53 (0.13), p < .001	0.52 (0.13), p < .001	0.53 (0.13), p < .001	0.51 (0.13), p < .001	0.51 (0.13), p < .001
Adolescent NDSS	2.30 (1.71), p = .178	2.36 (1.72), p = .171	2.51 (1.72), p = .144	2.30 (1.71), p = .178	2.13 (1.71), p = .213	2.15 (1.71), p = .210	2.47 (1.72), p = .150	2.42 (1.71), p = .157
Outcome: adolescent smoking frequency (intermediate aggregateb)
Heritable traita		−0.18 (0.46), p = .695	−0.02 (0.07), p = .754	0.36 (0.44), p = .421	0.14 (0.04), p = .002	0.26 (0.42), p = .539	0.89 (0.37), p = .018	0.63 (0.29), p = .032
Baseline smoking frequency	0.51 (0.07), p < .001	0.51 (0.07), p < .001	0.51 (0.07), p < .001	0.51 (0.07), p < .001	0.52 (0.07), p < .001	0.51 (0.07), p < .001	0.51 (0.07), p < .001	0.50 (0.07), p < .001
PTE	0.54 (1.06), p = .608	0.56 (1.05), p = .597	0.56 (1.06), p = .594	0.56 (1.05), p = .594	0.26 (1.05), p = .801	0.55 (1.06), p = .604	0.48 (1.05), p = .648	0.46 (1.05), p = .662
Maternal lifetime smoking	1.95 (0.69), p = .005	1.94 (0.69), p = .005	1.96 (0.69), p = .005	1.92 (0.69), p = .005	1.98 (0.68), p = .004	1.94 (0.69), p = .005	1.88 (0.69), p = .006	1.88 (0.70), p = .006
Baseline nicotine dependence	3.10 (0.725), p < .001	3.06 (0.74), p < .001	3.16 (0.74), p < .001	3.09 (0.73), p < .001	2.93 (0.72), p < .001	2.97 (0.76), p < .001	3.11 (0.72), p < .001	3.02 (0.72), p < .001
Outcome: adolescent nicotine dependence (intermediate aggregate measureb)
Heritable traita		−0.05 (0.04), p = .134	0.00 (0.01), p = .481	0.03 (0.03), p = .423	0.01 (0.00), p = .000	0.03 (0.03), p = .304	0.04 (0.03), p = .209	0.03 (0.02), p = .201
Baseline smoking frequency	0.02 (0.01), p < .001	0.02 (0.01), p < .001	0.02 (0.01), p < .001	0.02 (0.01), <0.001	0.02 (0.01), <0.001	0.02 (0.01), <0.001	0.02 (0.01), <0.001	0.02 (0.01), <0.001
PTE	−0.05(0.089), p = .539	−0.05 (0.08), p = .554	−0.055 (0.08), p = .569	−0.05 (0.08), p = .552	−0.07 (0.08), p = .350	−0.05 (0.078), p = .545	−0.05 (0.08), p = .513	−0.05 (0.08), p = .507
Maternal lifetime smoking	0.18 (0.05), p = .001	0.17 (0.05), p = .001	0.17 (0.05), p = .001	0.17 (0.05), p = .001	0.18 (0.05), p < .001	0.18 (0.05), p = .001	0.17 (0.05), p = .001	0.17 (0.05), p = .001
Baseline nicotine dependence	0.53 (0.06), p < .001	0.52 (0.06), p < .001	0.53 (0.06), p < .001	0.53 (0.06), p < .001	0.52 (0.06), p < .001	0.52 (0.06), p < .001	0.53 (0.06), p < .001	0.53 (0.06), p < .001

Items with significant p-values (p < .007, which reflects Bonferroni adjustment for seven comparisons) are shown in bold. PTE = Prenatal tobacco exposure; NDSS = Nicotine Dependence Syndrome Scale. B represents unstandardized coefficients.

^aHeritable personality traits (self-esteem, neuroticism, novelty-seeking, and impulsiveness) and heritable subjective first-time experiences (pleasurable, unpleasurable, and dizziness).

^bIntermediate aggregate measures are derived from all available timepoints after baseline and until age 18 (see text).

SEM results examining the role of each heritable personality trait and initial subjective experience trait are presented in Table 2. None of the heritable traits were directly associated with smoking frequency in young adulthood. All fit indices indicated good model fit. Like mother’s lifetime smoking, impulsiveness was positively associated with intermediate adolescent smoking frequency (B = 0.14, SD = 0.04, p = .002). This means that for every one-unit increase in BIS score, on average, adolescents smoked an additional 0.135 days of the month. Impulsiveness was also positively associated with intermediate adolescent nicotine dependence (B = 0.01, SD = 0.00, p < .001), such that with every additional unit on the BIS score, adolescents had a nicotine dependence score of 0.01 units higher on average. Both unpleasurable first experience (B = 0.90, SD = 0.37, p = .018) and dizziness (B = 0.63, SD = 0.29, p = .032) showed a trend with adolescent intermediate smoking frequency, but these were not significant after applying the Bonferroni correction (α = 0.007). Self-esteem, neuroticism, novelty-seeking, and pleasurable first experience were not significantly associated with smoking behavior or nicotine dependence either in adolescence or young adulthood.

Because impulsiveness was associated with adolescent intermediate smoking frequency, the possible mediation pathways of maternal lifetime smoking’s effect were examined next. The pathway between maternal lifetime smoking and impulsiveness was not significant (results not shown in the table; B = 0.56, SD = 0.70, p = .421), indicating that impulsiveness did not mediate maternal lifetime smoking’s effect on intermediate adolescent smoking frequency, but rather had an independent association. Thus, impulsiveness had significant pathways with young adult smoking frequency as well as intermediate adolescent smoking frequency and nicotine dependence but did not explain or mediate the effect of maternal smoking measures.

Discussion

The purpose of this study was to examine whether heritable traits explain previously observed pathways between maternal smoking measures (both PTE and lifetime smoking status) and offspring smoking behavior and nicotine dependence during adolescence and young adulthood.^5,14,15 The base model (with no heritable traits) replicated the previous findings from our group^5,14: Results show that of the heritable traits examined, impulsiveness was significantly associated with smoking behavior in young adulthood, but did not explain the effects of maternal smoking measures. In other words, the effect of maternal lifetime smoking and PTE on offspring smoking behavior does not appear to be explained by these heritable traits.

Impulsiveness was associated with both smoking frequency and nicotine dependence in adolescence, which in turn were associated with smoking frequency in young adulthood. This is consistent with previous literature, which found that individuals who smoke report higher impulsivity.³⁷ In addition, previous studies have shown that impulsivity was associated with current smoking³⁸ and that impulsivity was associated with cravings among individuals who smoke daily³⁹ but not individuals who smoke occasionally.⁴⁰ The current study also found a significant association between impulsivity and nicotine dependence independently of how much one smokes, though it remains unknown which specific dimensions of nicotine dependence, if any, drove this relationship.

Unpleasurable first experience showed a positive trend with adolescent smoking frequency but not adolescent nicotine dependence, though this was not significant after adjusting for multiple comparisons. The association between negative initial experiences and continued smoking has been shown in previous studies in various populations and ages.^41–43 These negative initial experiences often include dizziness, coughing, and nausea.^41,42 While this seems counterintuitive, unpleasant initial smoking experiences may indicate higher sensitivity to nicotine, which may promote continued smoking.⁴⁴

The current study also found that dizziness had a positive trend with adolescent smoking frequency, but not adolescent nicotine dependence, though this was not significant after adjusting for multiple comparisons. Several studies have found dizziness to be associated with continued smoking²³ and nicotine dependence.²⁴ Interestingly, dizziness can be associated with both positive and negative experiences for initial smoking experiences⁴²; however, the enjoyment of dizziness likely decreases with age.⁴⁵ The variability of dizziness at initial smoking experience warrants further research on how it impacts continued smoking and nicotine dependence, particularly among adolescents who may have more intense initial smoking experiences.

Self-esteem, neuroticism, novelty-seeking, and pleasurable first experience were not associated with any outcome variables (smoking frequency in young adulthood, intermediate adolescent smoking frequency, or intermediate adolescent nicotine dependence) independently of the other variables in the model. Though previous research has identified each of these as risk factors for smoking, it is important to note that these were tested in a model with many other variables. In other words, adolescent and young adulthood smoking behavior were better explained by the other variables in the model (ie, baseline smoking behavior and nicotine dependence, maternal smoking measures, and sex); therefore, the null findings here do not necessarily contradict previous work that examined simpler associations. Additionally, it is possible that some effects may manifest at other ages or stages of smoking not captured in this study.

As shown in previous models,^5,14,15 this study confirmed PTE’s direct effect on smoking frequency in young adulthood: those who have been exposed to tobacco prenatally, on average, smoked six additional days per month. The strength of this relationship was consistent across all SEMs and was statistically independent of the heritable traits examined. Thus, the heritable personality traits examined in this study did not explain the direct effect of PTE on smoking frequency in young adulthood. Thus, the mechanism by which PTE affects smoking behavior in young adulthood remains unclear.

Future studies should focus on environmental factors that may contribute to PTE’s effect on offspring smoking, such as parental rules and communication about smoking and parental attitudes towards smoking. For example, we have previously shown in a subset of this sample that the effect of maternal observed disclosure and elaboration of consequences for smoking on the likelihood that youth who experiment will continue to smoke was moderated by maternal smoking status.⁴⁶ Other future avenues of research could examine a potential behavioral teratologic effect detectable as early as infancy in behavioral disinhibition patterns that presage later smoking.⁴⁷ Large neurodevelopmental consortia with emphasis on prenatal substance exposure, such as ECHO⁴⁸ and the planned HEALthy Brain and Child Development (HBCD) to be linked with the ABCD study⁴⁹ (thus providing a rich characterization of the pathway from prenatal exposure to offspring smoking) provide an optimal design for the in-depth intergenerational, neurodevelopmental and outcomes requisite to elucidating this.

Finally, the current findings could inform tailored intervention strategies, especially with respect to the time-dependent associations. Namely, impulsivity and maternal lifetime smoking (and not PTE) are significant predictors of offspring smoking during adolescence. PTE (and not maternal lifetime smoking) and prior smoking behavior is a significant predictor of offspring smoking in young adulthood. Therefore, if screening efforts rely on these risk factors, it is important to consider the appropriate age range of the population being targeted for prevention efforts, and modify the variables being screened for to those that are significant during the corresponding age range.

Strengths

The data used in this study are from a robust longitudinal cohort, one of few which captures the earliest stages of smoking exposure. This study is unique in that it examined multiple hereditary traits as potential mechanisms that explain the effects of maternal lifetime smoking and PTE on offspring smoking. Structural equation modeling allowed for rigorous examination of simultaneous confounding and mediating pathways in examining the potential role of heritable personality and initial subjective experience traits in the effects of maternal smoking measures on offspring smoking outcomes in adolescence and young adulthood.

Limitations

This paper has several limitations. It is not clear how well these findings generalize to larger populations, based on study inclusion criteria (ie, oversampling adolescents who recently initiated and/or smoked at light levels) and characteristics of the sample (eg, high novelty-seeking scores). Individual variability may exist within significant pathways. We did not have data on preceding developmental pathways (eg, parental socialization) to adolescent and adult smoking. This study utilized self-reported data on smoking behavior, cumulative cigarettes, and mother’s concurrent and lifetime smoking; consequently, these behaviors may underreported.⁵⁰ These data were collected (2005–2015) before the upsurge in adolescent electronic nicotine device (e-cigarette use), and it is unknown whether maternal factors play an equivalent role adolescent use of electronic nicotine delivery system (ENDS) products. It is important to explore the specificity of vulnerability to the effects of nicotine versus the effects of using a particular delivery system. Finally, these results used observational data, and cannot definitively demonstrate causation, although SEM imposes a causal model on the data. In particular, socioeconomic status may explain some effects of maternal smoking measures, and additional research is needed to examine this potential confounding.

Conclusion

This study examined the effects of heritable personality or initial subjective experience traits within the pathway of maternal lifetime smoking and offspring smoking behaviors. Prevention efforts targeting high-risk adolescents and young adults are necessary to reduce smoking, and offspring of mothers who smoke are especially vulnerable. This line of research adds to current knowledge about the detrimental effects of maternal smoking, through examination of specific possible mechanisms through which this risk is conveyed. Given that none of the heritable traits examined in this study explain the effect of the maternal smoking measures we used, further research is necessary to identify the mechanism. Other possible explanations include a teratologic mechanism resulting from neurodevelopmental alterations of the fetal brain that predispose to smoking.¹³ Additionally, immediate next directions from our group include an investigation of potential environmental mechanisms such as parental rules and communications and access to cigarettes.

Funding

This work was supported by the National Cancer Institute (NCI; grant number P01CA098262), the National Institute for General Medical Sciences (NIGMS; grant number P20GM121341), and the National Institute on Drug Abuse (NIDA; grant number L40 DA042431) under the National Institutes of Health (NIH). LW was also supported by NIDA (grant number R34DA050266). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH, NCI, NIGMS, or NIDA.

Declaration of Interests

After the initial submission of this manuscript, AS became employed with Pinney Associates, which provides consulting services on tobacco harm minimization to JUUL Labs, Inc. The research presented here precedes this competing interest, and JUUL Labs, Inc. had no role in the conceptualization, design, analysis, interpretation, or presentation of data.