Does increasing cigarette use stabilize mood? A real-time investigation spanning 6 years of adolescence and young adulthood
Funding information: This research was supported by the National Cancer Institute of the National Institutes of Health (NIH) grant number P01CA098262 to R.J.M. The content is the sole responsibility of the authors and does not necessarily represent the official views of the NIH.
Abstract
Background and Aims
There is a longstanding assumption that cigarette smoking stabilizes mood. However, no studies have rigorously evaluated mood stability as people progress from occasional to regular use of tobacco cigarettes. This observational study thus tested two central questions: as smoking rates increase, (1) does the acute mood boost after smoking become more stable and (2) do background moods (i.e. mood levels outside of smoking) become more stable?
Design, Setting and Participants
Observational study of a cohort of n = 255 youth enriched for current smoking (mean age at baseline = 15.63 years, 52% female, 67% non-Hispanic White) recruited from 16 high schools in or near Chicago, Illinois, USA. They participated in up to 6 waves of ecological momentary assessment (EMA) spanning 6 years of their adolescence and young adulthood. During each week-long EMA wave, youth self-initiated reports of mood levels immediately before and after smoking. They also reported on background mood levels in response to random prompts ~5 times/day.
Measurements
Mixed-effects location scale (MELS) modeling tested the effects of within-person smoking rates on within-person variability in positive affect (PA) and negative affect (NA) levels modeled (1) from before to after smoking and (2) outside of smoking.
Findings
As smoking rates increased, on average, variability decreased by approximately 15–20% in the heightened PA (P < 0.01) and diminished NA (P < 0.01) changes from before to after cigarette use. Gender moderated the associations between smoking rates and mood variability during background random, non-smoking times: as smoking rates increased among boys only, on average, variability in background PA (P < 0.01) and NA (P < 0.01) decreased by around 10%.
Conclusions
As youth progress from occasional to more frequent cigarette use, their immediate mood boost after smoking appears to become more stable. Among boys only, background moods outside of smoking also appear to stabilize.
INTRODUCTION
The notion that nicotine regulates mood is central to many of the dominant theoretical models of dependence [1-3] as well as to people's own accounts of their reasons for ongoing use [4, 5]. These mood regulating effects may occur on both immediate and longer-term timescales. Most immediately, nicotine is associated with an acute mood boost including increased positive affect (PA) and decreased negative affect (NA) [6-8]. The magnitude of this boost is often small (e.g. half a point on a 10-point scale [6]), but may be meaningful precisely because of its nuance, perceptible to people as they smoke and allowing for fine-tuned adjustments to emotional experiences. To the extent that nicotine provides a reliable tool for regulating mood throughout the day, more frequent use over time could be associated with increasingly stable background mood (i.e. mood levels outside of smoking events)—albeit at the cost of greater dependence [9]. Indeed, escalating smoking rates may simultaneously stabilize mood and be necessary to stave of mood dysregulation related to withdrawal (e.g. greater anxiety and irritability [10, 11]). In line with this notion, there is a pervasive assumption among people who smoke that as their smoking increases, their background mood stabilizes. Such beliefs have been reflected over the decades in popular culture, for example, in a 1949 ad campaign from Lucky cigarettes that urged people to ‘Smoke a Lucky to Feel your Level Best!’ [12]. To date, however, we are not aware of studies that have rigorously tested the fundamental premise that as people progress from occasional to more frequent cigarette use, they experience increasingly stable mood. The present study aimed to address this gap.
Ecological momentary assessment (EMA), in which data are collected via brief self-reports as people go about their daily lives, is ideally suited to such investigation [13, 14]. For example, ratings of mood levels just before and after smoking over time could reveal if the acute mood boost from smoking becomes more reliable as smoking increases. Similarly, longitudinal sampling of mood levels at random intervals throughout the day could provide insights into whether increasing smoking rates are associated with increasing stability in background mood. Given that retrospective reports of both mood and common behaviors such as smoking are highly subject to recall bias, the completion of EMA in near-real time represents a substantial strength [13, 14]. Moreover, because EMA generates relatively large numbers of observations per person, it enables differentiation of within-person effects from between-person effects [15]. This offers a critical advantage given that emotion dynamics and substance use are within-person processes.
In terms of charting emotional experiences over smoking trajectories, the developmental period spanning adolescence and young adulthood represents a critical window. Most people who smoke regularly began during this developmental period, and an estimated 75% of high school students who smoke continue into adulthood [16, 17]. This same period marks a height in the onset of emotional distress (e.g. anxiety, depression) [18, 19]. The surge in distress appears to be driven in part by the large number of transitions that adolescents and young adults often navigate, such as starting and ending romantic relationships, school and/or new work and living arrangements [20, 21]. Against this unstable backdrop, it may not be surprising that many teenagers who experiment with cigarettes increasingly turn toward cigarettes' mood regulating effects as they progress through young adulthood. Moreover, heightened emotional distress has been associated with greater discomfort related to withdrawal [22, 23], suggesting that youth high in distress who begin smoking may be particularly likely to continue to stave off further discomfort.
Robust gender differences in emotional distress are also present during this period, with implications for youths' emotional responses to nicotine. Of particular relevance to this investigation, rates of most types of anxiety are about twice as high among girls as boys starting in middle childhood and continuing into adulthood [24, 25]. Anxiety, like depression, is characterized by high NA [26, 27]. A prominent symptom of this elevated NA that distinguishes anxiety from depression is anxious arousal, a physiological stress response mediated by the sympathetic nervous system [27, 28]. This means that as girls progress through adolescence and young adulthood, they on average experience greater sympathetic nervous system activity—such as elevated heart rate, respiration and muscle tension—than boys. This is important because nicotine acts in part on the sympathetic nervous system in exerting its subjective effects [9]. Nicotine absorbs rapidly into the bloodstream, triggering the near-immediate release of neurotransmitters that heighten positive emotions and inhibit negative emotions. These pathways lead to the acute mood boost consistently reported by people who smoke [9]. As a stimulant, however, nicotine is also liable to amplify anxious arousal, particularly among people whose sympathetic nervous systems are overactive [9, 29]. Although speculative, it follows that the more active a person's sympathetic nervous system, the more likely they may be to experience emotional dysregulation rather than stabilization after their acute mood boost subsides. This idea converges with data showing that not only is emotional distress, including anxiety, associated with greater withdrawal-related discomfort [22], but also that this effect may be stronger for girls than boys [30]. Put another way, although the immediate mood benefits from smoking may be relatively robust, the stabilizing effects of increased smoking on background moods may be less likely for females than males.
- Does the acute mood boost become more stable?
- Do background moods outside of cigarette use become more stable?
- We expected based on earlier work that the acute mood boost would generally become more reliable as smoking escalated (e.g. Hedeker and Mermelstein [6] and Hedeker et al. [7]).
- We expected, based on the gender differences in anxious arousal present during adolescence and young adulthood, that gender would moderate the relation between smoking rates and background mood [24, 25]: as smoking rates increased, variability in background mood levels would on average decrease among boys but not girls—suggesting that among boys only, in general, escalating cigarette use stabilized background moods.
METHODS
Participants
Participants were recruited into the Social–Emotional Contexts of Adolescent Smoking Patterns (SECASP) project, a longitudinal, observational study. Methodological details are available elsewhere [31]. Briefly, 9th- and 10th-grade students from 16 Chicago-area high schools were screened for cigarette smoking from 2005 to 2006; because e-cigarettes were not widely available, SECASP focused on combustible cigarettes. Screening data were used to form a longitudinal cohort enriched for high likelihood of smoking escalation based on self-report of recently smoking a cigarette (see Dierker and Mermelstein [31]). A total of n = 1263 adolescents provided assent, had parental permission and completed the baseline assessment. All participants were asked to complete questionnaire measures and semi-structured interviews at nine annual follow-up waves. Annual retention was high (e.g. 90% over the first 4 waves and 80% by wave 10).
A subset of participants was invited into a longitudinal EMA sub-study, the data from which were the focus of this report. The EMA sub-study spanned six waves occurring at baseline, 6 and 15 months and 2, 5 and 6 years (i.e. EMA waves 1–6). Recruitment occurred in two phases. First, n = 501 youth from the larger project were invited into EMA waves 1 to 4 at baseline; n = 461 provided assent, had parental permission and participated. Second, after receiving funds to extend the EMA sub-study, we recruited into EMA waves 5 to 6. We invited youth from the existing EMA cohort who still lived in the Chicago area and were likely to smoke (i.e. had self-initiated ≥1 smoking report during the first four EMA waves and reported smoking ≥1 time in the past 30 days on their 4-year follow-up questionnaire). We also invited youth from the larger project who were not previously in the EMA cohort, but otherwise satisfied the same criteria. In total, this yielded n = 367 invitations. From these, n = 307 youth provided assent, had parental permission and participated; n = 123 had participated previously in the EMA cohort. The analytic sample is comprised of the n = 255 youth who completed ≥2 random reports and self-initiated ≥2 smoking reports within any two of the six EMA waves. Approximately half the sample (56%) contributed EMA data at two waves and approximately half (44%) contributed data at three or more waves. EMA data collection occurred from 2005 to 2013, corresponding to mean participant ages of 15.63 (SD = 0.59) at baseline to 22.35 (SD = 0.74) at the final EMA wave.
Procedures
At each EMA wave, youth were asked to carry a handheld computer for 7 consecutive days. They were instructed to self-initiate EMA reports immediately after each cigarette and complete EMA reports when prompted at random intervals approximately 5 times/day. The Institutional Review Board at the University of Illinois Chicago approved study procedures.
Measures
Baseline demographics
Participants self-reported age, gender (male/female), race and ethnicity at baseline, among other constructs not included in this report.
Mood
In every EMA report, participants indicated how they felt ‘now’ using 10 adjectives each rated on a 10-point Likert scale. On the self-initiated reports, participants similarly rated how they felt ‘just before’ they smoked. Based on factor analysis and consistent with prior studies showing strong validity and reliability (e.g. Piasecki et al. [32] and Weinstein and Mermelstein [33]), composite scales were created for PA (average scores for ‘happy’, ‘relaxed’, ‘cheerful’, ‘confident’ and ‘accepted by others’) and NA (average scores for ‘frustrated’, ‘angry’, ‘stressed’, ‘irritable’ and ‘sad’). Higher scale scores indicated stronger emotions. Scores captured in the randomly-prompted reports reflected ‘background moods’ outside of smoking events.
Cigarette rate
Cigarette rate was measured by summing the self-initiated EMA smoking reports for each participant at each 1-week wave. For the wave-varying cigarette rate variable, both within-person and between-person effects were included in all models [34]. Both were centered around the means and scaled so that 1 unit = 10 cigarettes.
Data analyses
Two sets of mixed-effects location scale (MELS) models tested changes in (1) mood levels and variability from before to after cigarette use and (2) background mood levels and variability outside of cigarette use, both over EMA waves 1 to 6. MELS modeling extends the basic multi-level regression model by estimating the effects of covariates (listed below in the Results) on both mean and within-person variance [35, 36]. Each MELS model is comprised of two sub-models, run in parallel and conditional on each other, which regress the mean and within-person variance onto the covariates. The present models were constructed in SAS PROC NLMIXED, with observations nested in waves, nested in participants [6]. The PA and NA outcomes were run separately and gender (0 = female, 1 = male) was a moderator. All models included random subject mean and within-person variance effects, which were allowed to correlate, accounting for data correlations and heterogeneity. All models also included a random wave effect, allowing for correlations of responses from a participant within a timepoint, above and beyond overall correlations within participants. Model fit was evaluated using likelihood ratio tests to compare model complexity at each step. Mean-level effects are reported below as estimated coefficients. Variance effects are reported as variance ratios (VRs), calculated by exponentiating the estimated coefficients. VRs represent ratios of within-person variance in the outcome per unit change in the variable, with values >1, 1 and <1 indicating the variable is on average associated with increased, unchanged or decreased variability, respectively. Full information maximum likelihood estimation accommodated missing data [37]. The analysis was not pre-registered and the results should be considered exploratory.
RESULTS
The analytic sample (n = 255, 67% non-Hispanic White) averaged 15- to 16-years-old at baseline, with boys and girls about equally represented (Table 1). Over EMA waves 1 to 6, youth self-initiated 8243 (mean = 21.35, SD = 14.63) smoking reports and completed 31 975 (mean = 125.39, SD = 59.54) randomly-prompted reports. Compliance with random reporting was high, ranging from 69% to 87% across participants over the 6 waves. Average cigarette rates per week over waves 1 to 6 were 1.36 (SD = 2.66), 1.82 (SD = 3.10), 2.98 (SD = 4.51), 3.10 (SD = 4.46), 7.02 (SD = 6.57) and 6.48 (SD = 6.07), respectively. Table 2 presents descriptive statistics for changes in mood levels from before to after smoking and for background mood levels outside smoking at each EMA wave.
| Variable | Mean (SD) or n (%) |
|---|---|
| Age, y | 15.63 (0.59) |
| Female | 132 (51.75%) |
| Race/ethnicity | |
| Non-Hispanic White | 172 (67.45%) |
| Hispanic | 38 (14.90%) |
| Non-Hispanic Black | 31 (12.16%) |
| Asian/Pacific Islander | 5 (1.96%) |
| Other/unknown | 9 (3.53%) |
- Note: Sample (n = 255) at baseline. Gender was assessed as a binary construct.
| Change in mood levels with cigarette use (now-before) | Background mood levels outside of cigarette use | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Positive affect | Negative affect | Positive affect | Negative affect | ||||||
| Mean [95% CI] | Mean [95% CI] | Mean [95% CI] | Mean [95% CI] | ||||||
| Male | Female | Male | Female | Male | Female | Male | Female | ||
|
Wave 1 Total EMA: Male = 398 Female = 473 |
0.77 [0.46–1.08] | 0.74 [0.48–1.00] | −0.48 [−0.80 to −0.17] | −0.45 [−0.73 to −0.17] |
Wave 1 Total EMA: Male = 1710 Female = 2015 |
6.74 [6.46–7.02] | 6.52 [6.19–6.84] | 3.18 [2.87–3.50] | 3.96** [3.57–4.35] |
|
Wave 2 Total EMA: Male = 304 Female = 396 |
0.70 [0.41–0.98] | 0.40 [0.20–0.60] | −0.64 [−0.95 to −0.32] | −0.31 [−0.51 to −0.11] |
Wave 2 Total EMA: Male = 1288 Female = 1347 |
6.70 [6.41–7.00] | 6.80 [6.47–7.14] | 3.39 [2.95–3.83] | 3.69 [3.18–4.20] |
|
Wave 3 Total EMA: Male = 418 Female = 507 |
0.27 [0.01–0.53] | 0.40 [0.18–0.62] | −0.28 [−0.57 to 0.02] | −0.35 [−0.53 to −0.17] |
Wave 3 Total EMA: Male = 1162 Female = 1477 |
6.81 [6.38–7.25] | 6.76 [6.41–7.11] | 3.20 [2.78–3.62] | 3.66 [3.22–4.10] |
|
Wave 4 Total EMA: Male = 516 Female = 536 |
0.58 [0.36–0.80] | 0.27* [0.12–0.43] | −0.66 [−0.94 to −0.38] | −0.28* [−0.42 to −0.13] |
Wave 4 Total EMA: Male = 1402 Female = 1658 |
6.97 [6.67, 7.28] | 7.15 [6.84–7.47] | 3.21 [2.83–3.59] | 3.34 [2.95–3.72] |
|
Wave 5 Total EMA: Male = 1156 Female = 1227 |
0.50 [0.37–0.63] | 0.44 [0.29–0.58] | −0.35 [−0.50 to −0.20] | −0.36 [−0.47 to −0.24] |
Wave 5 Total EMA: Male = 3297 Female = 3618 |
7.12 [6.83–7.42] | 7.36 [7.09–7.63] | 3.15 [2.82–3.47] | 3.18 [2.86–3.49] |
|
Wave 6 Total EMA: Male = 974 Female = 1311 |
0.42 [0.29–0.55] | 0.34 [0.22–0.45] | −0.42 [−0.58 to −0.26] | −0.34 [−0.48 to −0.20] |
Wave 6 Total EMA: Male = 2728 Female = 3216 |
7.18 [6.85–7.51] | 7.25 [6.93–7.57] | 3.12 [2.78–3.47] | 3.09 [2.78–3.40] |
- Note: Significant gender differences at a given wave are indicated by *P < 0.05 and **P < 0.01. Positive and negative affect levels were rated on a 10-point scale. ‘Background mood levels’ were assessed in random (non-smoking) EMA reports. EMA waves 1–6 corresponded to baseline, 6 months, 15 months, 24 months, 5 years and 6 years, respectively. Total EMA = total number of EMA reports completed at a given wave. All results are from mixed models treating EMA observations nested within participants.
- Abbreviations: CI, confidence interval; EMA, ecological momentary assessment.
As cigarette use increases, does the acute mood boost become more stable?
The first MELS model examined changes in mood levels and variability from before to after smoking over time (Table 3). In the sub-model estimating mean changes, the intercept estimates indicated that for females at wave 0 with average smoking rates, there was an acute increase in PA and decrease in NA from before to after smoking (P < 0.01), on average. There was also a significant effect of within-person smoking rates on mean NA (P < 0.05), although this was an outlier among nonsignificant covariates within the mean sub-model (P > 0.05).
| Positive affect | Negative affect | |||||
|---|---|---|---|---|---|---|
| Estimate | 95% CI | P | Estimate | 95% CI | P | |
| Mean model | ||||||
| Intercept β0 | 0.497 | [0.35–0.65] | <0.0001 | −0.291 | [−0.40 to −0.18] | <0.0001 |
| Time (1 unit = 1 y) β1 | −0.026 | [−0.06 to 0.00] | 0.08 | −0.011 | [−0.03 to 0.01] | 0.30 |
| Male gender β2 | 0.140 | [−0.08 to 0.36] | 0.21 | −0.133 | [−0.29 to 0.03] | 0.11 |
| Between-person smoking rate β3 | −0.047 | [−0.12 to 0.03] | 0.23 | 0.030 | [−0.04 to 0.10] | 0.38 |
| Within-person smoking rate β4 | −0.045 | [−0.11 to 0.02] | 0.18 | 0.073 | [0.01–0.14] | <0.05 |
| Interaction effects | ||||||
| Gender × time β5 | −0.013 | [−0.06 to 0.03] | 0.56 | 0.031 | [0.00–0.06] | 0.06 |
| Gender × between-person smoking rate β6 | 0.011 | [−0.09 to 0.11] | 0.82 | 0.017 | [−0.07 to 0.11] | 0.71 |
| Gender × within-person smoking rate β7 | 0.006 | [−0.09 to 0.10] | 0.90 | −0.064 | [−0.15 to 0.02] | 0.12 |
| Variance model (exponentiated) | ||||||
| Intercept τ0 | 2.028 | [1.66–2.48] | <0.0001 | 1.825 | [1.45–2.30] | <0.0001 |
| Time (1 unit = 1 y) τ1 | 0.845 | [0.82–0.87] | <0.0001 | 0.876 | [0.85–0.90] | <0.0001 |
| Male gender τ2 | 1.338 | [1.01–1.79] | <0.05 | 1.247 | [0.89–1.74] | 0.11 |
| Between-person smoking rate τ3 | 0.867 | [0.73–1.03] | 0.10 | 0.811 | [0.66–1.00] | <0.05 |
| Within-person smoking rate τ4 | 0.863 | [0.81, 0.92] | <0.0001 | 0.789 | [0.74–0.85] | <0.0001 |
| Interaction effects | ||||||
| Gender × time τ5 | 0.965 | [0.92–1.01] | 0.11 | 0.980 | [0.94–1.03] | 0.39 |
| Gender × between-person smoking rate τ6 | 0.971 | [0.78–1.22] | 0.80 | 0.886 | [0.67–1.17] | 0.39 |
| Gender × within-person smoking rate τ7 | 1.023 | [0.93–1.12] | 0.63 | 1.022 | [0.93–1.13] | 0.67 |
- Note: Positive and negative affect levels were rated on a 10-point scale. P-values were based on Wald statistics (estimate/standard error ~ standard normal distribution). Variance model estimates were exponentiated to represent variance ratios; corresponding P-values reflected tests of variance ratios equaling 1.
- Abbreviations: CI, confidence interval; EMA, ecological momentary assessment.
In the variance sub-model, the intercept for PA and NA (P < 0.01) represented variance in the acute mood boost for females at wave 0 with average smoking rates. Supporting our first hypothesis, higher within-person smoking rates were associated with lower within-person variability in the enhanced PA and diminished NA from before to after smoking (P < 0.01). Specifically, as smoking rates increased by 10 cigarettes, variability decreased on average by approximately 14% for PA and 21% for NA. These effects emerged above and beyond covariates including the significant effects of time on PA and NA (P < 0.01), male gender on PA (P < 0.05), between-person smoking rate on NA (P < 0.05); and covariates in the mean sub-model.
As cigarette use increases, do background moods become more stable?
The second MELS model examined mean levels and variability in mood outside cigarette use (i.e. during random reports) (Table 4). Within the mean sub-model, the intercept indicated average levels of positive and negative background mood (P < 0.01) for females at wave 0 with average smoking rates. Covariates in the mean sub-model included the significant effects of time on PA and NA and male gender on NA (P < 0.01); between-person smoking rate on PA and within-person smoking rate on NA (P < 0.05); and the gender-by-time interaction on NA (P < 0.01).
| Positive affect | Negative affect | |||||
|---|---|---|---|---|---|---|
| Estimate | 95% CI | P | Estimate | 95% CI | P | |
| Mean model | ||||||
| Intercept β0 | 6.817 | [6.57–7.06] | <0.0001 | 3.779 | [3.47–4.09] | <0.0001 |
| Time (1 unit = 1 y) β1 | 0.106 | [0.05–0.16] | <0.0001 | −0.139 | [−0.20 to −0.07] | <0.0001 |
| Male gender β2 | −0.052 | [−0.41 to 0.30] | 0.77 | −0.604 | [−1.05 to −0.16] | <0.01 |
| Between-person smoking rate β3 | 0.302 | [0.07–0.54] | <0.05 | −0.130 | [−0.39 to 0.13] | 0.33 |
| Within-person smoking rate β4 | 0.089 | [−0.04 to 0.21] | 0.16 | −0.187 | [−0.34 to −0.03] | <0.05 |
| Interaction effects | ||||||
| Gender × time β5 | −0.033 | [−0.11 to 0.04] | 0.38 | 0.125 | [0.03–0.22] | <0.01 |
| Gender × between-person smoking rate β6 | −0.272 | [−0.59 to 0.04] | 0.09 | 0.038 | [−0.31 to 0.39] | 0.83 |
| Gender × within-person smoking rate β7 | −0.049 | [−0.24 to 0.14] | 0.60 | 0.122 | [−0.11 to 0.36] | 0.31 |
| Variance model (exponentiated) | ||||||
| Intercept τ0 | 2.296 | [2.01–2.62] | <0.0001 | 2.781 | [2.39–3.24] | <0.0001 |
| Time (1 unit = 1 y) τ1 | 0.930 | [0.92–0.94] | <0.0001 | 0.907 | [0.89–0.92] | <0.0001 |
| Male gender τ2 | 0.832 | [0.69–1.01] | 0.06 | 0.845 | [0.68–1.06] | 0.14 |
| Between-person smoking rate τ3 | 0.933 | [0.81–1.07] | 0.32 | 0.848 | [0.72–1.00] | <0.05 |
| Within-person smoking rate τ4 | 1.001 | [0.96–1.04] | 0.95 | 0.999 | [0.96–1.04] | 0.98 |
| Interaction effects | ||||||
| Gender × time τ5 | 1.023 | [1.00–1.05] | 0.06 | 1.021 | [1.00–1.05] | 0.10 |
| Gender × between-person smoking rate τ6 | 0.976 | [0.81–1.17] | 0.80 | 1.120 | [0.90–1.39] | 0.31 |
| Gender × within-person smoking rate τ7 | 0.910 | [0.85–0.97] | <0.01 | 0.895 | [0.84–0.96] | <0.01 |
- Note: Positive and negative affect levels were rated on a 10-point scale. ‘Background mood levels’ were assessed in random (non-smoking) EMA reports. P-values were based on Wald statistics (estimate/standard error ~ standard normal distribution). Variance model estimates were exponentiated to represent variance ratios; corresponding P-values reflected tests of variance ratios equaling 1.
- Abbreviations: CI, confidence interval; EMA, ecological momentary assessment.
In the variance sub-model, the significant intercept for PA and NA (P < 0.01) represented the variance in background, non-smoking mood levels for females at wave 0 with average smoking rates. The effects of within-person smoking rates on PA and NA were non-significant (P > 0.05), indicating no detectable effects of smoking rate on background mood stability among girls. Consistent with our second hypothesis, there was a significant effect for the interaction between gender and within-person smoking rate on variability in PA and NA (P < 0.01). This indicated a significant moderating effect: as smoking rates increased by 10 cigarettes, variability in background moods decreased, on average, for boys only by approximately 9% for PA and 11% for NA. This hypothesized finding emerged above and beyond covariates including the significant effects of time on PA and NA (P < 0.01); between-person smoking rates on NA (P < 0.05); and the mean sub-model covariates.
DISCUSSION
This study provided the first rigorous evaluation of the premise that as people progress from occasional to more frequent cigarette use, they experience increasingly stable mood—both in terms of stability in their acute mood boost following smoking and in background moods outside smoking. We examined within-person variability in real-time mood among a cohort of >250 youth over 6 years of their adolescence and young adulthood. As within-person smoking rates increased, on average, the acute mood boost became more stable across boys and girls. For boys but not girls, on average, background moods also became more stable as smoking rates increased. Over the 6-year study, average smoking increased from approximately once weekly to approximately once daily. The effects were, therefore, observed across rates that were relatively low but increasingly regular and might have been even stronger with greater smoking increases.
The first main finding, that the immediate mood benefits after smoking became more stable as smoking rates increased, replicated prior studies (e.g. Hedeker and Mermelstein [6] and Hedeker et al. [7]) while extending the observational window. As youths' smoking increased over 6 years, their average within-person variability in higher PA and lower NA following smoking decreased by approximately 15% to 20%. This suggested the acute mood boost generally became more reliable with greater use. That cigarettes functioned as an increasingly reliable tool for regulating mood comported with the larger idea that over time, increased smoking could stabilize background mood more generally (while also likely being required to stave off withdrawal) [10, 11].
Indeed, our second finding was that background moods became more stable with greater smoking rates, on average, but only among boys. Although smoking rates alone had no detectable effects on background mood stability, there was a significant moderating effect of gender. As rates increased among boys but not girls, on average, within-person variability in PA and NA outside smoking decreased by approximately 10%. This marks a novel contribution to the literature, but was anticipated based on established gender differences in anxiety. Essentially, girls are more likely than boys to experience heightened sympathetic nervous system activity including anxious arousal [24, 25]. Nicotine acts on the sympathetic nervous system to bring about its immediate mood benefits, which are experienced broadly across genders [9]. However, as a stimulant, nicotine is also liable to augment anxious arousal, particularly among people whose sympathetic nervous systems are more active. We, therefore, speculated that any longer-term stabilizing effects of cigarette use would more often be mitigated among girls than boys by heightened anxious arousal.
Although novel, these findings build on early work examining gender differences in mood stability related to cigarette escalation within the SECASP data. One analysis showed that among girls, higher within-person variability in negative mood at baseline predicted rapid smoking escalation by 15 months [33]. Another produced preliminary indications that changes in the stability and mean levels of negative mood over these 15 months varied as a function of gender and smoking stage [38]. These reports were limited, however, by small group sizes and the inclusion of only two EMA waves that spanned less than 1.5 years of youths' early smoking trajectories.
Confidence in the present findings, meanwhile, is reinforced by several methodological strengths. The data were collected from a relatively large cohort of youth over 6 years of adolescence/young adulthood. This represents the longest-running EMA cohort study of substance use of which we are aware, covering a critical developmental window [16]. The use of EMA minimized recall bias and maximized ecological validity [13, 14]. The EMA protocol combined self-initiated reports of smoking with random sampling, allowing for nuanced examination of emotional dynamics surrounding and outside of smoking events. Finally, MELS modeling differentiated within- from between-person effects in mood variability, allowing tests of our hypotheses on the within-person level while adjusting for numerous covariates within and between people [15, 35, 36].
The findings should nonetheless be considered in the context of design limitations. Although youth who smoke represent an important demographic—and the Chicago-area schools from which they were recruited were relatively diverse—the sampling strategy necessarily limited generalizability. The findings may not generalize to other groups including youth with differing smoking patterns or socio-economic backgrounds. Moreover, the observational design precluded causal inference. Participants were instructed to rate mood before and after each cigarette immediately after smoking, potentially introducing bias into the ‘before’ reports. It is possible some of the randomly-prompted reports coincided with cigarette use, and participants likely failed to report at least some cigarette events. However, given the large number of EMA observations and high compliance, and that random prompts were restricted from occurring soon after any event reports were made, it is unlikely these instances meaningfully impacted the results. In interpreting the findings, it should be noted that developmental changes may have affected mood stability in ways not fully addressed by the analyses, and there may have been other confounding variables such as social factors. Gender was assessed as a binary construct, leaving unknown how the findings would generalize across a spectrum of identities. Finally, e-cigarettes were not widely available at the time of data collection. Combustible cigarettes remain an important public health concern, however, and the findings likely have relevance to nicotine use more broadly. Even so, future studies should build on these results by examining e-cigarette events. Relatedly, the expanding legalization of cannabis in the United States has ushered in an explosion of nicotine-cannabis co-use [39]. Examination of how co-use impacts the stabilizing effects of nicotine on mood is, therefore, a worthwhile future direction. Underscoring this final point, recent findings suggested nicotine-cannabis co-use made the acute mood boost from nicotine less reliable, on average, compared with nicotine alone [40].
Despite its limitations, this study has important implications for understanding the reinforcing effects of nicotine as people progress from occasional to regular use. By charting the stabilizing effects of increased smoking on mood, the results lend empirical support to the subjective benefits described by many people who smoke (e.g. Kassel et al. [1], Khantzian, [2], Baker et al. [3], Fidler and West [4] and Piasecki et al. [5]). Although these benefits can come with serious costs including heightened risk for nicotine dependence, cancer and other illnesses, they are important to acknowledge as a valid part of why people continue smoking. Similarly, they speak to part of what people must give up when they quit cigarettes: an increasingly reliable tool for achieving an immediate mood boost and, at least among many males, stabilizing background mood.
AUTHOR CONTRIBUTIONS
Ashley D. Kendall: Conceptualization (equal); writing—original draft (lead); writing—review and editing (lead). Donald Hedeker: Conceptualization (equal); data curation (supporting); formal analysis (lead); funding acquisition (supporting); investigation (supporting); methodology (lead); project administration (supporting); validation (supporting); writing—original draft (supporting); writing—review and editing (supporting). Kathleen R. Diviak: Conceptualization (equal); data curation (lead); funding acquisition (supporting); investigation (supporting); project administration (supporting); validation (lead); writing—original draft (supporting); writing—review and editing (supporting). Robin J. Mermelstein: Conceptualization (equal); data curation (supporting); funding acquisition (lead); investigation (lead); methodology (supporting); project administration (lead); writing—original draft (supporting); writing—review and editing (supporting).
ACKNOWLEDGEMENTS
We thank Siu Chi Wong for assistance with data analysis.
DECLARATION OF INTERESTS
None.
Open Research
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from Robin J. Mermelstein ([email protected]) upon reasonable request.
