Data were drawn from a large longitudinal cohort of adolescents, the Adolescent Brain Cognitive Development (ABCD) study, assessed at multiple time points [19]. The assessments used in the present analyses took place between 2019 and 2024. Baseline Screen Time Questionnaire (STQ) assessments in the analytic sample were completed between March 2019 and October 2023, with subsequent UPPS (Urgency, Premeditation, Perseverance, and Sensation Seeking) and substance use follow-up assessments extending up to March 2024. The present analyses were restricted to participants with available STQ data [20] and UPPS impulsivity data [21], as well as information regarding nicotine use initiation [22]. To establish temporal ordering, participants who reported nicotine use initiation on or before the STQ assessment were excluded, yielding a nicotine-naive baseline cohort. Participants were further restricted to those with complete covariate data required for multivariable modeling. The final analytic sample consisted of adolescents followed prospectively for incident nicotine use initiation.

All models were adjusted for demographic covariates, including sex, age at baseline assessment, race, ethnicity, and household income. Income was included as a continuous variable. Analyses were conducted using a complete-case approach, with participants missing covariate data excluded from adjusted models.

Descriptive statistics were calculated for weekend texting screen time and negative urgency. Weekend texting screen time was assessed at the baseline STQ assessment. For each participant, negative urgency was obtained from the earliest available UPPS assessment taking place on or after the STQ assessment within the specified 2500-day eligibility window. Because the timing of the selected UPPS assessment varied across participants, including assessments taking place on the same day as the STQ, negative urgency was not treated as a uniformly subsequent assessment. Associations between baseline texting screen time and follow-up negative urgency were evaluated using linear regression models adjusted for demographic covariates.

Prospective associations between baseline texting screen time and incident nicotine use initiation were examined using Cox proportional hazards regression models [23]. The primary model estimated the total association between baseline texting screen time and time to nicotine use initiation, adjusting for demographic covariates. Participants were restricted to those who were nicotine naive at baseline. Time to event was defined as the number of days from the baseline texting assessment to the first recorded nicotine use onset date derived from longitudinal substance use data. Participants without nicotine use initiation were censored at their last available follow-up assessment. Because Cox proportional hazards models explicitly incorporated each participant’s observed time-to-event or censoring time, follow-up duration was not included as an additional covariate. We nevertheless examined the association between weekend texting screen time and follow-up duration to assess potential differential observation by exposure. To enhance interpretability, hazard ratios (HRs) were reported per 1-hour increase in weekend texting screen time and were additionally expressed for increases corresponding to 1 SD, 1 IQR, and 5 hours.

To examine whether negative urgency statistically accounted for part of the association between texting screen time and nicotine use initiation, a second Cox model included both texting screen time and negative urgency. Change in the texting coefficient following inclusion of negative urgency was used to quantify attenuation of the association. Because negative urgency was not uniformly assessed after STQ and before nicotine use initiation, these analyses were interpreted as associational rather than as establishing causal mediation.

As a sensitivity analysis addressing temporal ordering, we repeated the Cox analyses after excluding participants whose recorded nicotine use initiation occurred before or on the date of their selected UPPS assessment. This restriction ensured the sequence STQ assessment→UPPS negative urgency assessment→nicotine use initiation among participants experiencing the outcome. The time origin remained the baseline STQ assessment to preserve comparability with the primary survival analysis. Models used the same demographic covariates and family-level clustering as the primary analyses.

An exploratory product-of-coefficients analysis was additionally conducted on the log-hazard scale with nonparametric bootstrap CIs. Because the temporal ordering required for causal mediation was not consistently present in the primary sample, this analysis was interpreted only as a statistical breakdown of associations and not as an estimate of a causal indirect effect or causal proportion mediated.

Cox models incorporated clustering at the family level to account for nonindependence of observations. Proportional hazards assumptions were evaluated using Schoenfeld residual tests [24]. No violations of the proportional hazards assumption were detected. All analyses were conducted in R (R Foundation for Statistical Computing) [25] using the data.table and survival packages [26,27], along with custom code for bootstrap-based mediation analyses. Statistical significance was evaluated at a 2-sided α level of .05.

The ABCD study protocol was approved by the centralized Institutional Review Board (IRB) at the University of California, San Diego (160091), with participating study sites obtaining or relying on appropriate local IRB approval. Parents or legal guardians provided written informed consent, and participating youth provided assent. The present study involved secondary analysis of deidentified ABCD study data obtained through the National Institute of Mental Health Data Archive and was determined to be exempt from review by the University of California, San Diego IRB (815599). All study procedures were conducted in accordance with applicable ethical guidelines and the principles of the Declaration of Helsinki.

Of 8048 participants with available STQ and UPPS data, 771 individuals who reported nicotine use initiation on or before the STQ assessment were excluded to ensure a nicotine-naive baseline cohort. This yielded 7277 eligible participants. After restricting to individuals with complete exposure and mediator data and valid follow-up time, 5965 participants were included in the primary Cox proportional hazards analyses, during which 239 incident nicotine use initiation events were observed (Table 1).

In the primary analytic sample, participants had a median follow-up of 311 (IQR 232-366) days, with a maximum follow-up of 1407 days. A total of 28.7% (2827/9858) of participants had no follow-up Substance Use Interview assessment after the STQ, and 10.9% (1074/9858) had near-immediate follow-up (≤30 days). Follow-up duration was weakly inversely associated with weekend texting screen time (r=−0.044), indicating minimal differential follow-up by exposure (Table 2).

Weekend texting screen time and UPPS negative urgency scores were approximately normally distributed after cleaning. The negative urgency score had a mean of 8.36 (SD 2.43).

In linear regression models, greater weekend texting screen time was significantly associated with higher negative urgency scores (β=0.079; P<.001; Tables 3 and 4). Each additional hour of weekend texting screen time was associated with a modest but statistically robust increase in negative urgency, although the overall variance explained was small (R²=0.008).

Models were adjusted for demographic covariates (sex, age, race, ethnicity, and household income); full parameter estimates are provided in Table S1 in Multimedia Appendix 1.

In Cox proportional hazards models, greater weekend texting screen time was significantly associated with increased risk of incident nicotine use initiation. Each additional hour of weekend texting screen time was associated with a 7% higher hazard of initiation, corresponding to an HR of 1.07 with a 95% CI of 1.04 to 1.11. Model discrimination was modest, with a concordance of 0.617. There was no evidence of violation of the proportional hazards assumption.

The magnitude of this association was clinically meaningful when scaled to larger differences in exposure. A 5-hour increase in weekend texting screen time corresponded to approximately a 40% higher hazard of nicotine use initiation. When standardized, a 1-SD increase in weekend texting screen time was associated with a 21% higher hazard of nicotine use initiation (HR=1.21, 95% CI 1.10‐1.31).

When UPPS negative urgency was included in the model, it independently predicted nicotine use initiation, with each 1-point increase associated with a 19% higher hazard. The association between weekend texting screen time and nicotine use initiation was modestly reduced after adjustment for negative urgency (HR=1.06 per hour), although the magnitude of this attenuation was small. The persistence of the texting association after adjustment indicates that negative urgency statistically accounted for only a portion of the observed association. Because negative urgency was not uniformly assessed before nicotine use initiation, this attenuation should not be interpreted as evidence of a causal indirect effect. In standardized models, the adjusted for negative urgency effect of weekend texting screen time remained statistically significant (HR=1.17, 95% CI 1.06‐1.28), whereas a 1-SD increase in negative urgency was associated with a 49% higher hazard of nicotine use initiation (HR=1.49, 95% CI 1.29‐1.73).

Exploratory bootstrap analyses yielded a statistically significant product-of-coefficients estimate for the association involving weekend texting, negative urgency, and nicotine use initiation (Table 5). On the log-hazard scale, the product estimate was 0.014 (95% CI 0.008‐0.022). The standardized product estimate was 0.0365 (95% CI 0.0201‐0.0561). Because negative urgency did not uniformly precede nicotine initiation in the primary analytic sample, these estimates are presented as statistical decompositions rather than causal indirect effects, and the corresponding proportion should not be interpreted as the proportion of the total effect causally mediated through negative urgency.

Because negative urgency did not uniformly precede nicotine use initiation in the primary analytic sample, we conducted a sensitivity analysis restricted to participants for whom the temporal sequence of baseline STQ assessment, UPPS negative urgency assessment, and nicotine use initiation was preserved. Of the 239 nicotine use initiators in the primary sample, 109 had their selected UPPS assessment before nicotine use onset, whereas 130 had recorded nicotine use onset before the selected UPPS assessment. Exclusion of the latter participants yielded 5835 participants and 109 events; after restriction to complete covariate data, the adjusted sensitivity models included 5345 participants and 96 events.

Weekend texting remained associated with subsequent nicotine use initiation in the temporally ordered sample (HR=1.11, 95% CI 1.06‐1.16; P<.001). After adjustment for negative urgency, the association remained significant (HR=1.10, 95% CI 1.05‐1.16; P<.001), and negative urgency was independently associated with nicotine use initiation (HR=1.11, 95% CI 1.02‐1.21; P=.01). Inclusion of negative urgency attenuated the weekend texting log-hazard coefficient by 6.8%. There was no evidence of violation of the proportional hazards assumption (global P=.83 for both models).

Sensitivity analyses using a fallback definition of nicotine use initiation based on the first observed nicotine use at follow-up visits yielded consistent results with slightly larger effect sizes. In these models, both weekend texting screen time and negative urgency remained strongly associated with nicotine use initiation, and model discrimination was higher, reflecting increased event capture in this framework.

Overall, the association between weekend texting and nicotine use initiation was robust across alternative outcome definitions.

In this longitudinal analysis of nicotine-naive early adolescents, greater weekend texting screen time was associated with increased risk of subsequent nicotine use initiation. This association persisted after adjustment for demographic covariates and remained statistically significant after adjustment for negative urgency. Weekend texting was also positively associated with negative urgency, and negative urgency was independently associated with nicotine use initiation. Adjustment for negative urgency produced modest attenuation of the texting-initiation association, suggesting that the two measures capture partially overlapping information relevant to early nicotine use risk.

Importantly, negative urgency was not uniformly measured before nicotine use initiation in the primary analytic sample, precluding a strong temporal or causal mediation interpretation. However, in a sensitivity analysis restricted to participants with temporally ordered STQ, UPPS, and nicotine use onset assessments, both weekend texting and negative urgency remained associated with subsequent nicotine use initiation. These findings support the robustness of the observed associations while indicating that negative urgency accounts for only a modest portion of the texting-initiation relationship.

Although the per-hour association was modest, larger exposure differences were associated with clinically meaningful increases in nicotine use initiation risk. A 5-hour increase in weekend texting corresponded to an approximately 30% to 40% higher hazard of initiation, with modest attenuation after accounting for negative urgency. Consistent with this, standardized models indicated that a 1-SD increase in weekend texting screen time was associated with a 21% higher hazard of initiation. Given the highly skewed distribution of texting behavior, such increments represent meaningful shifts in exposure and suggest that high-intensity digital engagement may carry substantive risk.

Negative urgency demonstrated a robust association with use initiation. A 1-SD increase in negative urgency corresponded to approximately 50% higher hazard of nicotine use onset, consistent with extensive literature linking urgency-related impulsivity to early substance use and escalation. These findings reinforce the central role of affect-driven impulsive behavior in adolescent nicotine vulnerability.

Prior research has demonstrated associations between overall screen time and adolescent substance use, including smoking and vaping [8,9]. However, much of this literature has relied on cross-sectional designs, aggregated heterogeneous forms of screen behavior, or failed to establish temporal ordering between exposure and initiation [10,17]. By restricting analyses to nicotine-naive adolescents and using prospective time-to-event modeling, the present study strengthens inference regarding temporal sequencing.

Our focus on weekend texting should not be interpreted as evidence that weekend exposure is uniquely associated with nicotine use initiation or that weekday texting and other interactive digital behaviors are unimportant. Rather, weekend texting represents the specific measure of discretionary, socially interactive digital engagement examined in the present study. Future studies directly comparing weekday and weekend texting and other screen modalities will be necessary to determine whether the observed association is specific to weekend messaging or reflects broader patterns of digital social engagement.

Few studies have examined specific interactive forms of screen engagement such as texting and messaging. Text-based communication may differ from passive screen behaviors (eg, television viewing) in that it involves real-time social interaction, emotional salience, and peer reinforcement [15,28]. Consistent with this distinction, prior literature has suggested that screen time requiring interaction with others, such as texting, may be more strongly associated with vaping than more passive forms of screen use, such as watching videos [29]. Such environments may amplify social contagion effects or increase exposure to peer norms related to substance use [30,31].

The observed role of negative urgency aligns with theoretical models emphasizing affect regulation and impulsivity as key drivers of adolescent risk behavior. Youth high in negative urgency may be more likely to engage in maladaptive coping strategies, including substance use, particularly in emotionally charged contexts [11,12,14]. Digital communication environments may heighten emotional reactivity, social comparison, or interpersonal stress, thereby interacting with urgency-related traits [16,28]. Our findings provide empirical support for this conceptual framework.

The attenuation of the texting coefficient after inclusion of negative urgency suggests that negative urgency statistically accounts for a modest portion of the association between weekend texting and nicotine use initiation. This interpretation was supported by the temporally ordered sensitivity analysis, in which both weekend texting and negative urgency remained associated with subsequent nicotine use initiation. However, these findings should not be interpreted as establishing negative urgency as a causal mediator or as indicating that texting episodes directly provoke nicotine use through acute negative affect. Rather, greater digital social engagement and negative urgency may represent correlated behavioral and individual-level characteristics associated with vulnerability to early nicotine use initiation [32,33].

This pattern suggests that additional mechanisms, as well as sources of residual confounding, likely contribute. These may include peer network exposure and social influence processes [30], digital advertising and product marketing [34,35], sensation-seeking tendencies not fully captured through negative urgency [36], and sleep disruption associated with high digital engagement [37]. Future work should examine these pathways directly, potentially incorporating peer network analyses, ecological momentary assessment, or digital trace data.

The large and diverse ABCD cohort provides an important opportunity to examine risk factors for nicotine use initiation beginning in early adolescence. However, participants were approximately 10 years old at the baseline texting assessment, and nicotine use initiation during the observed follow-up therefore represents relatively early onset. The findings should consequently not be generalized to nicotine use initiation across the full adolescent age range, particularly initiation occurring during mid- or late adolescence, when prevalence is substantially higher. Instead, the present results identify behavioral and individual difference correlates of comparatively early nicotine use initiation. Replication across later developmental periods will be important. Because assessments used in the present analyses spanned 2019 to 2024, the observed associations reflect the nicotine product and digital media environments during that period; both environments have continued to evolve.

Several limitations warrant consideration. First, texting screen time was self-reported and may be subject to recall bias or social desirability effects. Objective digital use metrics would strengthen future studies. Second, although the prospective design improves temporal inference, residual confounding cannot be excluded. Unmeasured variables such as peer substance use, parental monitoring, or mental health symptoms may partially account for observed associations. Additionally, other screen modalities, particularly social media use, may confound this association as adolescents who text heavily on weekends may also have greater exposure to peer modeling or nicotine-related content online.

Third, the timing of the negative urgency assessment varied across participants. For approximately half of the analytic sample, the UPPS assessment took place within 1 day of the STQ assessment, whereas for others, the selected UPPS assessment took place substantially later. Moreover, among some participants who initiated nicotine use, the selected UPPS assessment took place after the recorded nicotine use onset date. This timing precludes interpretation of the primary negative urgency analyses as demonstrating causal mediation. We therefore interpret negative urgency as statistically accounting for part of the observed association rather than as an established causal intermediary. Reassuringly, sensitivity analyses restricted to participants for whom STQ assessment preceded UPPS assessment and UPPS assessment preceded nicotine use initiation yielded consistent associations for both weekend texting and negative urgency. Nevertheless, studies with prospectively scheduled and clearly separated exposure, mediator, and outcome assessments are needed to establish mediation.

Future research should extend this work in several ways. First, incorporating objective digital use data would allow for more precise quantification of exposure patterns and timing. Second, examining additional impulsivity facets and affect regulation constructs may clarify which personality dimensions most strongly link digital engagement to substance use risk. Third, integrating peer network measures could test whether social contagion mediates part of the association. Fourth, longitudinal models that evaluate reciprocal influences among digital engagement, impulsivity development, and substance experimentation would deepen mechanistic understanding.

From a prevention standpoint, these findings suggest that adolescents with high digital engagement and elevated negative urgency may benefit from interventions targeting emotion regulation and coping skills. Digital literacy efforts may also help reduce exposure to substance-related content in online social environments.

In a longitudinal cohort of nicotine-naive early adolescents, greater weekend texting screen time was associated with a higher hazard of subsequent nicotine use initiation. Negative urgency was independently associated with initiation and modestly attenuated the texting-initiation association. These associations persisted in a sensitivity analysis requiring temporal ordering of texting assessment, negative urgency assessment, and nicotine use initiation. Although these findings do not establish negative urgency as a causal mediator, they suggest that digital engagement and affect-driven impulsivity represent partially overlapping correlates of early nicotine use risk. Future longitudinal studies with clearly separated assessment of exposure, potential mediators, and nicotine use onset are needed to clarify the mechanisms underlying these associations.