Nicotine use and Psychoticism: A Longitudinal Co-Twin Control Analysis

Abstract

Nicotine use, particularly in adolescence, has been linked to increased psychosis risk in observational studies; however, the causal status of these associations remains unclear. We analyzed data from 3,836 twins from two longitudinal community samples with repeated nicotine use measures from adolescence into adulthood and a dimensional measure of psychosis-proneness (PID-5 Psychoticism) in adulthood. We used individual-level mixed-effects models and co-twin control analyses that decompose associations into between-pair (familial) and within-pair (controlling for shared genetic and environmental factors) components, covarying sex, zygosity, cohort, age, and co-occurring substance use. Adolescent nicotine use predicted adult psychoticism at the individual level (β=0.09, 95% CI [0.06, 0.12]). Co-twin control analyses revealed significant between-pair effects (β=0.10, 95% CI [0.07, 0.14]) but nonsignificant within-pair effects (β=0.04, 95% CI [−0.02, 0.11]), suggesting familial confounding. Effects attenuated after accounting for adulthood nicotine use, suggesting longitudinal associations are attributable to continuity of nicotine exposure. Prospective co-twin control models over four years in adulthood also yielded null within-pair effects (β=0.02, 95% CI [−0.07, 0.11]). In contrast, cross-sectional adulthood associations persisted at the within-pair level (β=0.12, 95% CI [0.07, 0.17]), consistent with proximal rather than enduring effects. Taken together, associations between adolescent nicotine use and adulthood psychoticism appear largely attributable to shared familial factors and continuity of nicotine use into adulthood. Similarly, longitudinal analyses across two adult timepoints provide no evidence for enduring prospective causal effects. However, adult nicotine use shows cross-sectional associations with psychoticism, independent of familial confounding, suggesting potential causal effects over shorter-term timescales warrant further investigation.

General Scientific Summary:

Although numerous studies have suggested that adolescent nicotine use increases the risk of psychosis in adulthood, this large twin study challenges that assumption by showing these associations are likely due to shared familial factors and continuity of nicotine use into adulthood rather than direct causal effects. Similarly, results of longitudinal analyses during adulthood suggest no evidence of prospective causal effects over multi-year periods. In contrast, adulthood nicotine use showed cross-sectional associations with psychoticism independent of familial confounding. These findings suggest prevention efforts should target shared risk factors underlying both nicotine use and psychosis (e.g., childhood adversity), though the complex relationship in adulthood warrants further investigation.

Introduction

Nicotine use remains a significant public health concern, particularly among adolescents and young adults. Although the prevalence of traditional cigarette smoking has declined in many high-income countries, the rise of e-cigarettes has led to increased nicotine exposure among youth since their introduction into the market in the mid-2000s (Perikleous et al., 2018). In the United States, nicotine use is widespread and often co-occurs with other substance use (Wang et al., 2021). Recent estimates from 2023 indicate that 27.9% (4.39 million) of high school students reported lifetime use of a nicotine product, with 12.6% (1.97 million) identified as current users – most commonly of e-cigarettes (Birdsey et al., 2023). Among adults, 18.7% reported current nicotine use in 2021, with cigarettes being the most common product (Cornelius et al., 2023). These statistics underscore the persistence of nicotine use and highlight the need for continued research delineating its long-term consequences, including potential increased risk for psychopathology.

Among the potential health outcomes, the relationship between nicotine and psychotic symptoms and disorders has drawn considerable interest. Recent reviews and meta-analyses report strong and consistent associations between nicotine use and psychosis, with dose-response effects suggesting that higher intensity and prolonged use increase psychosis risk (Gurillo et al., 2015; Hunter et al., 2020; Myles et al., 2012; Scott et al., 2018). Historically, the leading explanation for this association was reverse causation – that individuals experiencing psychosis use nicotine to alleviate symptoms or medication side effects, a phenomenon known as “self-medication” (Kumari & Postma, 2005). Although nicotine has been linked to potential therapeutic effects, such as reducing antipsychotic side effects and improving cognitive deficits in individuals with psychosis (Quigley & MacCabe, 2019), evidence for the self-medication hypothesis remains limited. Longitudinal studies increasingly suggest that nicotine use often precedes the onset of psychosis, challenging the notion that individuals initiate use primarily to self-medicate after symptoms emerge (Myles et al., 2012; Scott et al., 2018).

Alternatively, nicotine may directly contribute to psychosis. Supporting this hypothesis, a review of eight prospective, longitudinal studies found positive, dose-response associations between nicotine use and psychosis in the majority (Scott et al., 2018). This potential causal link may be explained by neurochemical mechanisms. Nicotine alters signaling in dopaminergic, cholinergic, and glutamatergic neurotransmitter systems, leading to prolonged increases in dopamine release — a key mechanism implicated in the pathogenesis of psychosis (Yuan et al., 2015). Thus, the relationship may emerge as a result of two complementary processes: transient, immediate changes in dopamine function that increase susceptibility to psychotic experiences (proximal effects; Maskos et al., 2005; Wittenberg et al., 2020), and lasting neurobiological alterations (e.g., changes in receptor expression) that alter neural connectivity and contribute to later psychosis risk (enduring effects; Subramaniyan & Dani, 2015), particularly when use occurs in sensitive periods like adolescence (Smith et al., 2015). If such causal relationships were confirmed, targeted interventions to reduce nicotine use may have the potential to lower psychosis prevalence in the population.

A third possibility is the shared liability hypothesis, which posits that the high prevalence of nicotine use among individuals at risk for psychosis reflects common underlying genetic and environmental vulnerabilities. Nicotine use is associated with established risk factors for psychosis, including childhood adversity (Banzer et al., 2017; Elliott et al., 2014; Jaffee, 2017), lower socioeconomic status (Ávila-Burgos et al., 2023), concurrent use of other substances (e.g., cannabis) (Agrawal et al., 2011), and genetic predispositions (Fisher et al., 2024; Prantzalos & Turner, 2022). However, a recent review of the relationship between nicotine and psychosis concludes that shared genetic liability cannot fully explain the association (Quigley & MacCabe, 2019), leaving unresolved whether nicotine has a direct causal role or whether the link reflects confounding from co-occurring substance use or shared environmental influences.

Although existing studies suggest a potential causal relationship, attempts to control confounding factors have been limited to statistical adjustment of measured variables. This approach has known limitations, including the inability to account for unmeasured confounders, as well as residual confounding due to measurement error. These challenges highlight the need for more informative studies to clarify the association between nicotine and psychosis. One powerful method for addressing confounding is the co-twin control design, which compares monozygotic (MZ) and dizygotic (DZ) twins who differ in their nicotine exposure. This approach allows examination of the effects of nicotine use while simultaneously controlling for all genetic and environmental factors—measured and unmeasured—shared between twins (McGue et al., 2010). If nicotine plays a causal role in psychosis, twins with greater nicotine exposure should show increased psychosis risk relative to their lesser-exposed co-twins. Conversely, if no such differences are observed, mutual genetic and shared environmental factors likely drive the association. Prior work from our group applied this approach to examine adolescent cannabis exposure and adulthood psychoticism, finding associations were attributable to shared familial factors rather than causal effects, even among genetically vulnerable individuals (Schaefer et al., 2021). Whether nicotine—which has distinct neurobiological mechanisms and patterns of use compared to cannabis—has causal effects on psychoticism remains an open question.

To date, few studies have investigated the nicotine–psychosis link using co-relative or co-twin control designs. Existing studies of this type suggest a small independent effect of nicotine but find the association is largely attributable to shared familial factors (Barkhuizen et al., 2019; Kendler et al., 2015; Lyons et al., 2002). However, these studies are constrained by methodological limitations, particularly their reliance on retrospective and coarse self-report measures of nicotine use (e.g., no/light/heavy smoking) and psychotic symptoms. Such measures are prone to reporting and recall biases, which reduce their accuracy. Additionally, the one study that employed a longitudinal design, finding a higher risk of schizophrenia for the smoking twin relative to the non-smoking cotwin in monozygotic pairs, had inconsistent follow-up periods between sexes (Kendler et al., 2015).

The present study aims to address these limitations by examining associations between adolescent nicotine exposure and adulthood psychosis-proneness in two longitudinal community twin samples maintained by the Minnesota Center for Twin Family Research (MCTFR) and the Institute for Behavioral Genetics (IBG). Unlike prior co-twin control studies, these cohorts include repeated, gold-standard self-report and interview-based assessments of nicotine use during adolescence, reducing concerns about both retrospective bias and coarse measures. Consistent with accumulating evidence indicating that psychosis is a dimensional phenomenon with similar etiological influences across clinical and subthreshold manifestations (Guloksuz et al., 2019), psychosis-proneness was assessed using the Psychoticism scale of the Personality Inventory for DSM–5 (PID-5), administered in adulthood. In addition to leveraging our prospective twin design to test for evidence of a potential causal relationship between adolescent nicotine exposure and adult psychoticism—and to examine whether these associations persist after controlling for adult nicotine use—we also tested whether adult nicotine use exerts a more proximal, independent effect on psychoticism. By leveraging a prospective twin design, repeated nicotine use assessments, and a dimensional psychosis-proneness measure, this study provides a rigorous test of the contribution of nicotine use to psychosis risk.

Method

Participants and Procedures

We analyzed data from 3,836 twins from two longitudinal community twin samples maintained by: the Minnesota Center for Twin Family Research (MCTFR; n = 2,055) and the Institute for Behavioral Genetics at the University of Colorado Boulder (IBG; n = 1,781). Specifically, data come from three subsamples from MCTFR (older, younger, and enrichment cohorts) and two subsamples from IBG (Longitudinal Twin Study, Community Twin Study); detailed recruitment procedures and sample characteristics have been described previously (Corley et al., 2019; Wilson et al., 2019). Briefly, participants were born between 1972–1994 and recruited through publicly available birth records. Although participant age at intake (birth – 11 years) and assessment schedules varied between sites, both followed a similar longitudinal design, with assessments conducted prospectively through adulthood using highly similar measures. A final, coordinated assessment wave was conducted across both sites during adulthood (ages 24–49; 2018–2021). In the present study, primary analyses used data from assessments targeting late adolescence in each cohort, as well as the coordinated adulthood wave.

The final analytic sample was 58.2% female and 56.3% MZ pairs. The average age at the adolescence wave was 17.8 (SD = 1.0) and at the adulthood wave was 35.6 (SD = 5.3). The majority (91.8%) of participants were White and 5.3% of the sample reported Hispanic ethnicity, consistent with the birth cohorts and states from which these participants were drawn.

Measure

Nicotine Use.

To evaluate adolescent nicotine exposure effects, we harmonized measures across sites to create a '30-day nicotine use frequency' variable. In Minnesota, nicotine use was assessed in an interview as days per typical month in which participants used nicotine products in the last 12 months (range: 0–30). In Colorado, nicotine use was assessed in an interview as a continuous number of days used in the last 180 (range: 0–180). To harmonize nicotine use frequency across sites, the Colorado “number of days used in the past 180” was divided by 6 to approximate the number of days used per typical month. In adulthood, nicotine use was assessed uniformly via interview as days per typical month in which participants used nicotine products in the last 12 months (range: 0–30).

To assess the robustness of our findings, we examined alternative indicators of adolescent nicotine exposure. First, lifetime use was coded as a binary variable indicating whether participants reported ever using nicotine in adolescence. Second, age of onset captured the age in years at which participants first used a nicotine product among those who reported lifetime use, to assess developmental timing effects. Third, we created an ordinal nicotine use intensity variable representing average cigarettes per day (or equivalent tobacco product, e.g., chewing tobacco; Elkins et al., 2018): 0 = none; 1 = 1–2 cigarettes; 2 = 3–9; 3 = 10–19 (at least half a pack); 4 = 20–39 (one to two packs); and 5 = 40+ (two or more packs). This final measure was available for those in the Minnesota twin sample only and captured the typical intensity of nicotine consumption beyond frequency of use.

Psychoticism.

Psychoticism was assessed across all participants in the coordinated adulthood assessment using the abbreviated Psychoticism scale from the Personality Inventory for DSM–5 (PID-5; Maples et al., 2015). The Psychoticism scale is composed of three facet scales: Unusual Beliefs & Experiences (e.g., “Sometimes I can influence other people just by sending my thoughts to them”), Eccentricity (e.g., “Others seem to think I'm quite odd or unusual”), and Perceptual Dysregulation (e.g., “It’s weird, but sometimes ordinary objects seem to be a different shape than usual”). The PID-5 Psychoticism scale has demonstrated validity in differentiating individuals with and without psychosis and distinguishing psychotic disorders from other psychiatric conditions (Bastiaens et al., 2019; Longenecker et al., 2020). To ensure data quality, the self-report measure included validity checks such as simple arithmetic questions, and only participants with valid responses were included in analyses (40 excluded). Internal consistency was good for the total Psychoticism scale at the coordinated adulthood assessment (α = 0.87).

Covariates.

Adolescent alcohol and cannabis use were assessed concurrently with nicotine use across both sites. In Minnesota, alcohol and cannabis use was assessed through a combination of interview and self-report for frequency in the last 12 months; self-report measures incorporated validity checks (embedded questions about nonexistent substance) to ensure attentive and valid responses, and only valid measures were included in analyses (20 excluded). In Colorado, both were measured via interview as days used in the last 180 days. We harmonized these measures across sites to create 30-day frequency variables, applying the same methodology used for nicotine use. During the coordinated adult assessment across both sites, 30-day alcohol and cannabis use frequency were assessed via interview.

Statistical Analyses

We first conducted preliminary analyses, including descriptive statistics and twin correlations for study variables. All statistical models were fit using the lmer and lmerTest packages, which estimate linear mixed models with p-values (Kuznetsova et al., 2017). Prior to fitting statistical models, all continuous variables (nicotine use frequency, PID-5 Psychoticism scores and facet scales, and covariates) were standardized (M = 0, SD = 1) within the analytic sample. Estimates from all models are reported as standardized betas, with those for nicotine exposure reflecting the standard deviation increase in PID-5 Psychoticism scores for each standard deviation increase in nicotine use frequency. Statistical significance was set at α = .050 (corrected for multiple comparisons using Hochberg's method; Blakesley et al., 2009).

Individual-Level Models

To estimate the phenotypic association between adolescent nicotine use and adulthood psychoticism, we first fit mixed effects models, capturing individual-level effects adjusting for family-level dependency. Specifically, models included a fixed effect of adolescent nicotine use, fixed effects covariates including sex, zygosity, cohort membership, and age at the time of outcome assessment, and a family-level random intercept term to account for the nested twin data. Separate models were fit for the psychoticism total score and for each of the three facets as dependent variables.

We then examined whether observed associations persisted after controlling for two key types of twin-specific confounds. First, we added past-month frequency of alcohol and cannabis use as additional fixed effects covariates to examine whether associations persisted after controlling for concurrent use of alcohol and cannabis in adolescence, given high rates of co-occurring substance use (Wang et al., 2021). This analysis aimed to determine whether the association between nicotine use and psychoticism was directly related to nicotine use or potentially explained by broader substance use patterns during adolescence. Second, we added adulthood nicotine use frequency as an additional fixed effect covariate to the original model to examine whether the associations with adolescent nicotine use remained significant after controlling for adulthood nicotine use, which was assessed contemporaneously with psychoticism. This analysis was designed to examine whether observed effects of adolescent nicotine use reflected unique developmental risk for adulthood psychoticism, above and beyond the influence of ongoing nicotine use in adulthood. To ensure that inclusion of these covariates did not introduce problematic multicollinearity, we examined variance inflation factors (VIFs), which quantify how strongly predictors correlate with one another, with higher values indicating a greater likelihood that collinearity could bias estimates or inflate standard errors.

Twin Models

Co-twin Control Models.

We estimated co-twin control models for each set of analyses described in the previous section to isolate the effects of adolescent nicotine exposure on psychoticism after controlling for shared genetic and rearing environmental influences, which may confound this association. These models decompose each individual twin's nicotine exposure into between-pair and within-pair components (Begg & Parides, 2003; McGue et al., 2010). The between-pair component represents the twin pair's mean nicotine use, whereas the within-pair component represents each twin's deviation from their pair's mean. The between-pair effect in this parameterization largely captures genetic and shared environmental liability toward nicotine use among reared-together twins, whereas the within-pair effect provides a more isolated test of the exposure effect, capturing non-shared environmental causes. A significant, positive between-pair effect indicates that twin pairs with higher average nicotine use relative to other twin pairs also have higher average psychoticism, reflecting familial-level associations. A significant, positive within-pair effect indicates that, in twin pairs differing in nicotine use, the greater using twin also has higher psychoticism, consistent with a causal effect of nicotine on psychoticism.

We conducted co-twin control analyses in the full sample using the same three-step approach as the individual-level models: initial models including only sex, zygosity, cohort membership, and age at the time of outcome assessment as fixed effect covariates, adjusted models for concurrent use of other substances, and adjusted models for adulthood nicotine use assessed contemporaneously with psychoticism. Then, we repeated the full three-step co-twin control analysis separately for MZ and DZ twin pairs, excluding opposite-sex twin pairs to simplify interpretation of results. This stratification addresses whether associations differ by level of genetic similarity, with MZ twins sharing 100% of their genetic material and DZ twins sharing approximately 50%. Similar within-pair associations across MZ and DZ twins would suggest that observed effects are not substantially confounded by shared genetics.

Biometric Twin Models.

Bivariate biometric variance decomposition models were used to partition the variance in nicotine use and psychoticism into genetic and environmental sources. These models supplement the primary cotwin control analyses by estimating not only the genetic and environmental contributions to each phenotype individually, but also the extent to which genetic and environmental influences are shared between adolescent nicotine use and adulthood psychoticism. The genetic correlation quantifies the degree to which the same genetic factors influence both traits, while the environmental correlations quantify the overlap in environmental influences—with the nonshared environmental correlation capturing the unique environmental factors experienced by one twin but not the other (including differential nicotine exposure). Twin models typically partition variance into additive genetic (A), shared environmental (C), and nonshared environmental (E) components. However, when MZ correlations substantially exceed twice the DZ correlations, this pattern indicates nonadditive genetic effects (dominance; D) rather than shared environmental influences, warranting ADE models (additive genetic + dominance genetic + nonshared environmental) instead of ACE models.

Biometric models were run using OpenMx in R (Boker et al., 2021), with parameters estimated via Cholesky decomposition. Confidence intervals were obtained through bootstrap resampling of twin pairs with replacement (k = 1000 replications). All models included sex and age as covariates, as unadjusted mean differences could otherwise bias variance component estimates (McGue & Bouchard, 1984).

Sensitivity Analyses

Alternate Exposure Measures.

To assess the robustness of our findings, we conducted parallel individual-level and co-twin control analyses using three alternative adolescent nicotine exposure measures: binary lifetime use, continuous age of nicotine onset, and ordinal cigarettes per day. Lifetime use analyses included the full analytic sample, whereas onset age analyses were restricted to participants who reported lifetime use and provided valid onset data (n = 1,906). Cigarettes per day analyses were limited to the Minnesota sample due to data availability (n = 1,997). These alternative measures address distinct theoretical questions: lifetime use examines whether any adolescent nicotine exposure has enduring effects; onset age tests developmental timing sensitivity; and cigarettes per day evaluates dose-response relationships. Together, these analyses provide a more comprehensive assessment addressing how different dimensions of adolescent nicotine exposure relate to adulthood psychoticism, thereby also maximizing comparability with prior research.

Stratified Analyses.

We conducted stratified analyses by study site and sex to explore whether the associations between adolescent nicotine exposure and adulthood psychoticism differed across these subgroups.

Transparency and Openness

We report how we determined our sample size, all data exclusions (if any), all manipulations, and all measures in the study. Data cleaning, analyses, and visualization were conducted in R Version 4.4.1. The above analysis plan was preregistered on April 21, 2025, and is available at: https://osf.io/afh4q/. In addition to the preregistered analyses, we conducted follow-up analyses examining nicotine use in adulthood, which were not preregistered and should be considered exploratory. Analytic code and output will be made available at the same link upon publication (Sloan et al., 2025), though study data are not publicly available due to privacy concerns and limitations of the informed consent provided by participants at each assessment.

Results

Preliminary Analyses

Of all participants with adolescent data, 64.1% had adulthood psychoticism data, and those missing the outcome reported significantly greater nicotine use during adolescence (p < .001). Descriptive data for our measures of substance use and psychoticism within the analytic sample are presented in Table 1, with twin correlations between all continuous variables presented in Supplemental Table S1. For nearly all traits, monozygotic twin correlations were more than twice the corresponding dizygotic twin correlations, a pattern consistent with nonadditive genetic influences and suggesting minimal shared environmental contributions.

Table 1. Descriptive data for substance use and PID-5 Psychoticism.

		Minnesota (N=2055)		Colorado (N=1781)		Combined Analytic Sample (N=3836)
Time		M (SD)	Range	M (SD)	Range	M (SD)	Range
Adolescence
	Nicotine Use Frequency	5.89 (10.92)	0.00 – 30.00	3.91 (9.36)	0.00 – 30.00	4.97 (10.28)	0.00 – 30.00
	Alcohol Use Frequency	1.30 (2.94)	0.00 – 14.00	1.27 (2.86)	0.00 – 30.00	1.28 (2.90)	0.00 – 30.00
	Cannabis Use Frequency	1.08 (4.36)	0.00 – 25.00	1.25 (4.78)	0.00 – 30.00	1.16 (4.57)	0.00 – 30.00
Adulthood
	PID-5 Psychoticism	0.31 (0.39)	0.00 – 2.50	0.35 (0.41)	0.00 – 2.58	0.33 (0.40)	0.00 – 2.58
	Unusual Beliefs & Experiences	0.25 (0.43)	0.00 – 2.75	0.29 (0.47)	0.00 – 3.00	0.27 (0.45)	0.00 – 3.00
	Eccentricity	0.55 (0.68)	0.00 – 3.00	0.64 (0.73)	0.00 – 3.00	0.59 (0.70)	0.00 – 3.00
	Perceptual Dysregulation	0.11 (0.27)	0.00 – 2.50	0.11 (0.27)	0.00 – 2.25	0.11 (0.27)	0.00 – 2.50
	Nicotine Use Frequency	5.15 (10.71)	0.00 – 30.00	4.17 (9.76)	0.00 – 30.00	4.70 (10.29)	0.00 – 30.00
	Alcohol Use Frequency	6.59 (7.85)	0.00 – 30.00	6.83 (8.16)	0.00 – 30.00	6.70 (8.00)	0.00 – 30.00
	Cannabis Use Frequency	2.26 (6.94)	0.00 – 30.00	2.92 (7.80)	0.00 – 30.00	2.57 (7.36)	0.00 – 30.00

Note. All values are calculated based on the full analytic sample (N=3,836; twins with PID-5 Psychoticism data and adolescent nicotine exposure). Frequency scores reflect a harmonized measure of number of days used per typical month (e.g., 30-day frequency).

Nicotine use in adolescence was highly prevalent in our analytic sample, with approximately half of the participants (50.1%) endorsing history of use. Among those who endorsed nicotine use in adolescence, approximately one-quarter (24.8%) endorsed daily use. Our analytic sample included 1,543 complete twin pairs, and 983 pairs (63.7%) had no difference in adolescent nicotine use frequency (i.e., same exact frequency reported by both twins). Among pairs reporting different nicotine use frequency, the mean absolute difference was 3.92 days (SD = 8.75), with the large SD indicating that large differences account for a substantial portion of discordant pairs. Moreover, 86 pairs (15.4% of discordant twin pairs) showed maximal discordance, with one twin reporting no current nicotine use and the other reporting maximum use (i.e., 30 of the past 30 days). For our primary outcome, 349 twin pairs (22.6%) had identical PID-5 Psychoticism scores, while 500 pairs (32.4%) differed by more than one standard deviation. Overall, 439 twin pairs (28.5%) had differences in both nicotine use and psychoticism. These patterns suggest that although many twin pairs had similar nicotine use and psychoticism scores, a substantial proportion showed meaningful differences, allowing within-family comparisons that enable us to isolate the unique effects of individual nicotine exposure while accounting for shared familial liability.

Is Greater Nicotine Exposure in Adolescence Associated with Greater Psychoticism in Adulthood?

Results from individual-level models are visualized in Figure 1, with the corresponding statistical values presented in Supplemental Table S2. Consistent with prior research, more frequent adolescent nicotine use was associated with higher scores on the Psychoticism scale and its facets in adulthood (βs=.05-.11, ps ≤.001). These associations remained significant after adjusting for concurrent alcohol and cannabis use in adolescence (Supplemental Table S3); variance inflation factors for predictors were near 1, indicating no evidence of multicollinearity. However, after controlling for adulthood nicotine use assessed contemporaneously with Psychoticism, effect sizes for adolescent use were attenuated and largely nonsignificant; in contrast, adulthood nicotine use showed positive associations with all Psychoticism outcomes (Supplemental Table S4). Variance inflation factors were near 1, again suggesting no evidence of multicollinearity.

Figure 1. Associations between adolescent and adulthood nicotine use and adulthood psychoticism.

Note. Effects of adolescent and adulthood nicotine use on adulthood psychoticism are shown. Results are from three models: a base model (adjusted for age, sex, zygosity, cohort), a second model adjusting for concurrent alcohol and cannabis use, and a third model adjusting for nicotine use at the alternate time point (e.g., adolescent effects are adjusted for adulthood nicotine use). Points represent standardized effect size estimates, and error bars denote 95% confidence intervals.

Does Adolescent Nicotine Exposure Predict Greater Adulthood Psychoticism Independent of Shared Familial Factors, Consistent with a Potential Causal Effect?

Results from co-twin control analyses are also visualized in Figure 1, with the corresponding model estimates and p-values presented in Supplemental Table S2. In the full sample, between-pair associations between adolescent nicotine exposure and adulthood psychoticism were statistically significant (βs=.06-.13, ps≤.005), whereas within-pair associations were smaller and nonsignificant (βs=.03-.06, ps>.05). Results were similar across DZ and MZ pairs in zygosity-stratified analyses (Supplemental Table S5), and complementary bivariate ADE models showed a significant additive genetic correlation (r_A = .63, 95% CI [.11, 1.00]), but no significant genetic dominance or, critically, non-shared environmental correlations (r_D and r_E, respectively) (Supplemental Figure S1). Together, these findings are consistent with the interpretation that the observed phenotypic association between adolescent nicotine use and adulthood psychoticism reflects largely a shared genetic liability. The absence of significant within-pair effects in the co-twin control analyses, coupled with the near-zero r_E estimate in the bivariate ADE model, provides little evidence for a causal non-shared environmental effect of adolescent nicotine use on adulthood psychoticism.

Sensitivity Analyses

Individual-level models revealed endorsing a history of lifetime nicotine use, earlier nicotine use initiation, and greater cigarettes per day in adolescence were associated with higher adulthood Psychoticism scores, even after adjusting for concurrent alcohol and cannabis use in adolescence (Supplemental Table S6). Co-twin control models using these measures generally showed significant between-pair effects and nonsignificant within-pair effects, or—in the case of lifetime nicotine use—effects that became nonsignificant after adjusting for co-occurring adolescent substance use.

Individual-level associations between adolescent nicotine use and adulthood psychoticism were consistently positive across all site - and sex-stratified models, although the magnitude and statistical significance of these associations varied. For co-twin control analyses, similar patterns emerged across site- and sex-stratified models, with positive between-pair effects and smaller, consistently nonsignificant within-pair effects. Detailed results from these stratified analyses are available in Supplemental Tables S7-S8.

To examine whether findings were robust to the conditional nature of nicotine use frequency on initiation, we conducted supplemental bivariate ADE models separately examining initiation in the full sample and frequency among ever-users only (n=1,921). These analyses revealed that initiation and frequency show distinct patterns of association with psychoticism; importantly, the ever-user restricted frequency model — most directly relevant to our focus on intensity of use — were consistent with the primary results, suggesting our findings are not artifacts of conflating never- and ever-users (Supplemental Figures S2-S3).

Exploratory Analyses of Adulthood Nicotine Use

Since adulthood nicotine use was positively associated with all psychoticism outcomes and largely accounted for the adolescent nicotine-psychoticism associations, we explored its independent relationship with psychoticism. Results are visualized in Figure 1, with the corresponding statistical values presented in Supplemental Table S9. Individual-level models showed that more frequent adulthood nicotine use was associated with higher scores on the Psychoticism scale and its facets (βs=.09-.16, ps <.001), even after adjusting for concurrent alcohol and cannabis use in adulthood.

Co-twin control models revealed significant between-pair effects for psychoticism and its facets (βs=.11-.18, ps <.001) and significant within-pair effects for psychoticism and two of its facets (βs=.11-.12, ps <.001), all surviving adjustment for adult alcohol and cannabis use (Supplemental Table S10). MZ twin analyses—fully controlling for shared genetics—showed significant within-pair effects for overall psychoticism and one facet (Supplemental Table S11), suggesting adulthood nicotine use is associated with psychoticism independent of familial confounding, consistent with potential causation. Site- and sex-stratified analyses showed consistent patterns (Supplemental Tables S12-S13). Bivariate genetic modeling showed a small non-shared environmental correlation (r_E=.09, 95% CI [.02, .16]) between adulthood nicotine use and adulthood psychoticism, supporting possible causal effects, though mutual environmental influences and reverse causation cannot be excluded (Supplemental Figure S4).

To examine whether findings were robust to the conditional nature of nicotine use frequency on initiation, we conducted supplemental bivariate ADE models separately examining initiation in the full sample and frequency among ever-users only (n=2,533), with individual-level and co-twin control results for lifetime nicotine use reported in Supplemental Table S14. The ever-user restricted frequency model was consistent with the primary results, showing a significant rE (.11, 95% CI [.03, .19]; Supplemental Figure S5). In contrast, initiation models showed no significant genetic or environmental correlations (Supplemental Figure S6), again suggesting that initiation and frequency show distinct patterns of association with psychoticism, and that the nonshared environmental overlap may be specific to frequency of use rather than initiation.

Given the absence of evidence supporting a causal longitudinal association from adolescence to adulthood, alongside evidence of robust cross-sectional associations in adulthood, we conducted additional exploratory longitudinal analyses within adulthood. Specifically, we leveraged a young-adult assessment wave with identical measures of nicotine use and psychoticism (α = 0.95) available for a subset of participants in the MCTFR sample (n = 1,227; ages 22–39; mean interval between assessments = 3.9 years, SD = 0.9). These longitudinal analyses were designed to test whether nicotine use predicts subsequent psychoticism within adulthood and, conversely, whether psychoticism predicts subsequent nicotine use, allowing us to evaluate potential reverse causation.

We first examined models predicting adulthood psychoticism as a function of young-adult nicotine use. Individual-level models showed that greater nicotine use in young adulthood was prospectively associated with higher scores on the Psychoticism scale (β = .07, p = .013) and two of its facets, even after adjusting for corresponding young-adult psychoticism scores, and consistent with results for prospective individual-level models for adolescent nicotine use and adult psychoticism (Supplemental Table S15). However, co-twin control models revealed near-zero, nonsignificant within-pair effects for all outcomes, also consistent with results for adolescent nicotine use and adult psychoticism. We then examined models predicting adulthood nicotine use as a function of young-adult psychoticism. At the individual level, higher young-adult Psychoticism scores (β = .05, p = .044) and two of its facets were prospectively associated with increased nicotine use in adulthood, even after adjusting for young-adult nicotine use (Supplemental Table S16). Again, co-twin control models revealed near-zero, nonsignificant within-pair effects. Taken together, these findings indicate prospective associations in both directions at the individual level, with little evidence for causal effects in either direction.

Discussion

Results from this longitudinal study advance understanding of the relationship between nicotine use and psychoticism in three ways. First, we showed that greater nicotine exposure in adolescence is associated with greater psychoticism in adulthood. Second, co-twin control analyses suggest these prospective associations may reflect confounding due to shared familial liability and co-occurring substance use rather than a causal effect of nicotine exposure. Consistent with this interpretation, associations between adolescent nicotine use and adulthood psychoticism were substantially attenuated after accounting for contemporaneous adult nicotine use, indicating that early nicotine exposure may be linked to later psychoticism primarily through continuity of use into adulthood rather than a direct long-term effect. Third, exploratory analyses revealed that contemporaneous nicotine use in adulthood demonstrated a more robust association with psychoticism than adolescent use, with significant within-pair associations persisting after accounting for co-occurring adult substance use, consistent with potentially causal effects. In contrast, but consistent with results for adolescent nicotine use and adult psychoticism, exploratory longitudinal analyses conducted in a subsample with available data at two timepoints during adulthood found no evidence of prospective within-pair effects in either direction. Together, this pattern suggests that short-term causal effects and/or other nonshared, time-varying environmental factors may drive the association during adulthood, whereas the prospective association between adolescent, as well as adult, nicotine use and adulthood psychoticism is largely attributable to shared familial liability and developmental continuity of nicotine exposure.

Our findings extend results from previous studies of the association between adolescent nicotine use and later psychotic outcomes in several ways. First, consistent with prior meta-analyses and systematic reviews, we observed that greater nicotine exposure in adolescence was associated with greater psychoticism in adulthood, aligning with evidence suggesting a dose-response effect whereby higher intensity and prolonged nicotine use elevate psychosis risk (Gurillo et al., 2015; Hunter et al., 2020; Myles et al., 2012; Scott et al., 2018). Second, our co-twin control analyses indicate that this association may primarily reflect confounding by shared familial liability and co-occurring substance use rather than a direct causal effect of nicotine exposure. This aligns with previous co-twin control studies largely demonstrating that familial factors largely account for the observed associations, though we found no evidence to support the small independent effect of nicotine that has been previously reported (Barkhuizen et al., 2019; Kendler et al., 2015; Lyons et al., 2002). These conclusions held across site- and sex-stratified models, further supporting the confounding interpretation in adolescence. Although these results do not rule out the possibility that the association between adolescent nicotine use and adulthood psychosis may vary depending on vulnerability factors we did not consider, the null results from our co-twin control models and consistent findings across stratified models suggest that, if these especially vulnerable individuals exist, they likely make up a very small part of the overall population. Moreover, the attenuation of adolescent nicotine effects after adjustment for adult nicotine use suggests that previously reported long-term associations may, in part, reflect simple persistence of nicotine use from adolescence into adulthood—with individuals who use more in adolescence also using more in adulthood—rather than enduring neurodevelopmental effects of adolescent use itself.

These findings align with and extend previous work from our group examining associations between adolescent cannabis use and psychoticism using similar prospective, co-twin control methods in overlapping samples (Schaefer et al., 2021). That study similarly found no evidence for causal effects of adolescent cannabis exposure on adulthood psychoticism, suggesting associations were attributable to shared familial factors. The consistency of null causal findings for both nicotine and cannabis—despite their distinct pharmacological profiles and both being strongly implicated in psychosis through observational research—suggests that prospective associations between adolescent substance use and later psychoticism may primarily reflect shared familial vulnerabilities rather than substance-specific developmental effects. Collectively, these results underscore the value of using genetically informed designs to disentangle familial confounding from causal effects across multiple substances and developmental stages; future work jointly modeling nicotine, cannabis, and alcohol use within a quasi-experimental framework could allow formal decomposition of substance-specific versus general polysubstance pathways to psychosis.

The pattern of associations in adulthood mirrors that observed for adolescent nicotine use in some ways but diverges in others, revealing a complex temporal dynamic. Exploratory longitudinal analyses examining two timepoints during adulthood found no evidence of prospective causal effects in either direction over multi-year periods, consistent with findings for adolescent use. However, robust cross-sectional associations between contemporaneous nicotine use and psychoticism in adulthood—including significant within-pair associations among monozygotic twins—suggest a different mechanism may operate at shorter timescales. These cross-sectional adult within-pair associations (overall within pair β ≈ .12) remained significant even after accounting for co-occurring substance use, and bivariate genetic models indicated a large additive genetic correlation alongside a small but nonzero nonshared environmental correlation. This pattern—significant cross-sectional associations but null prospective effects—suggests that adult nicotine use is most strongly associated with concurrent levels of psychoticism rather than predicting future changes. Such a pattern is compatible with either extremely short-lived causal effects that do not persist or accumulate over time (e.g., acute nicotine/withdrawal effects or short-term self-medication) and/or nonshared, time-varying confounding factors that fluctuate contemporaneously with both nicotine use and psychoticism (e.g., sleep disruption, acute stress, other substances). Notably, these findings in a nonclinical community sample extend the literature on clinical research showing elevated smoking rates among individuals with psychotic disorders (Han et al., 2023; Rogers et al., 2024) and align with emerging evidence of associations between nicotine use and subclinical psychotic traits (Bourgin et al., 2020; Rep et al., 2023; Yousef et al., 2024). Together with the attenuation of adolescent effects after accounting for adult nicotine use, this pattern reinforces that contemporaneous adult exposure—rather than early developmental effects—may be the key driver of the nicotine-psychoticism association.

The discrepancy between robust cross-sectional within-pair associations and null longitudinal findings requires careful consideration of alternative explanations. First, nicotine's effects on psychoticism may be primarily contemporaneous rather than persistent, consistent with neurobiological evidence of nicotine inducing transient alterations in dopamine signaling (Smith et al., 2015; Subramaniyan & Dani, 2015); our multi-year assessment intervals may be too long to capture such short-term effects. Second, unmeasured nonshared environmental factors that vary contemporaneously with both nicotine use and psychoticism—such as acute stress, social isolation, or co-occurring mental health symptoms—may account for the cross-sectional associations even after controlling for familial factors. Third, the cross-sectional adulthood findings may reflect cumulative exposure effects – rather than short-term causal effects – as chronic nicotine use has been linked to neuropsychological impairments including cognitive impulsivity and reduced cognitive flexibility (Conti et al., 2019).

This study has several strengths. Using a dimensional measure of psychosis-proneness as an outcome rather than relying on categorical diagnoses of psychotic disorders allowed us to detect subtle variations in psychotic traits across the general population, capturing the full spectrum of psychosis vulnerability. Further, our use of a continuous measure of psychosis-proneness (PID-5 Psychoticism) alongside a continuous measure of nicotine use maximized statistical power to detect true effects. Additionally, our use of prospective, repeated assessments of nicotine use using gold-standard, interview-based measures reduces the well-known limitations of retrospective reporting and minimize measurement error that could bias within-pair estimates. Finally, by examining nicotine–psychoticism associations at different developmental periods and across different timescales within the same longitudinal twin sample, we provide a rigorous test addressing whether nicotine use contributes to psychosis risk across different stages of the lifespan, controlling for shared familial vulnerability as well as key measured nonshared covariates (i.e., other substance use).

Nevertheless, our study has several limitations that warrant consideration. First, although site- and sex-stratified results were consistent, our sample of predominantly White twins from Colorado and Minnesota born in the last quarter of the 20^th century may not generalize to other populations; expanding research to more contemporary and diverse samples would improve generalizability and examine whether results replicate when considering more recently popularized nicotine products (e.g., e-cigarettes). Second, selective attrition may have biased our results, as adolescents without available adulthood data reported greater nicotine use than adolescents with adulthood data; while associations are generally robust to selective non-response, as long as some individuals with extreme scores remain in the data (Gustavson et al., 2019), this raises the possibility that our findings underestimate the true association between adolescent nicotine use and adulthood psychoticism. Third, assessing psychoticism at a single time point in adulthood for most of our sample—and only one other timepoint in a subset of the sample—precludes a rigorous examination of developmental trajectories and comprehensive testing of simple versus reverse causation with maximal statistical power. Longitudinal studies with repeated assessments across developmental stages and shorter timescales would provide a clearer picture of the temporal relationship between nicotine use and psychoticism, including whether critical periods of vulnerability exist. Relatedly, reliance on a single self-report measure of psychosis is an additional limitation; a single instrument cannot fully disambiguate psychosis-proneness from related constructs (e.g., neuroticism, subclinical conversion-like phenomena), and future work should prioritize incorporating multiple complementary measures for a better triangulation of the nicotine-psychosis association. Finally, the co-twin control design, though powerful, is vulnerable to measurement error that can disproportionately bias within-twin-pair estimates compared to unpaired associations (Frisell et al., 2012), cannot fully rule out nonshared environmental confounds, and does not account for potential second-hand nicotine exposure of the lesser-using twins in discordant pairs.

Implications for research follow directly from these considerations. First, researchers studying this topic should prioritize designs that resolve short-timescale dynamics (e.g., ecological momentary assessment with daily nicotine exposure, craving/withdrawal, stress, sleep, and psychotic-like experiences) and apply within-pair fixed-effects or change-on-change models to test contemporaneous coupling. Second, future studies should include a broad range of plausible nonshared confounders and evaluate attenuation of adult within-pair effects when these are included. Third, randomized cessation/reduction trials or quasi-experimental leverage (e.g., policy shocks) paired with high-frequency outcomes may aid the identification of putative short-term effects.

Implications for policy and practice follow from the finding that, in adults, higher current nicotine exposure co-occurs with higher current psychoticism within families (including MZ twins), supporting clinical screening for nicotine use among individuals reporting psychotic-like experiences and integration of cessation or reduction support with the expectation of possible short-term symptom benefits. At the same time, our longitudinal results argue against claims that reducing nicotine will necessarily yield durable reductions in psychoticism over multi-year horizons. Clinical messaging should therefore emphasize near-term symptom management and broader health benefits. Finally, despite the lack of evidence for a causal pathway from adolescent nicotine use to adulthood psychoticism in these data, adolescent nicotine use prevention remains a public health priority given its many other established harms.

In conclusion, our findings suggest that widely reported associations between adolescent nicotine use and increased risk of psychosis in adulthood likely reflect confounding due to shared familial liability and continuity of nicotine exposure into adulthood, rather than a causal effect of adolescent nicotine use itself. Similarly, longitudinal analyses examining two timepoints during adulthood found no evidence of prospective causal effects in either direction, suggesting the relationship may be more complex than simple unidirectional causation. However, contemporaneous nicotine use in adulthood shows cross-sectional associations independent of familial confounding. These results underscore the need for both a lifespan and dimensional perspective when evaluating nicotine's role in psychosis outcomes, as well as more intensive longitudinal designs with shorter assessment intervals to better capture the temporal dynamics of this relationship. By further clarifying these associations, we can develop more targeted and effective strategies for reducing the burden of psychotic illness.