Depression as a moderator of the prospective relationship between mood and smoking

Abstract

Objective.

Affective features of depression are uniquely involved in the depression-smoking relationship, and it follows that smokers with depression are likely to use cigarettes to alleviate negative affect. However, most ecological momentary assessment (EMA) studies demonstrate no relationship between mood and smoking, in general. Conversely, a small number of experimental studies suggest there is an association between mood and smoking and that the relationship is dependent on levels of depression. Researchers have yet to examine the impact of depression on the relationship between mood and smoking using EMA methodology. Accordingly, the aim of this study was to explore the relationship between depression, mood, and ad-lib smoking among adults in “real-time”.

Methods.

Participants included 96 adult daily smokers (53% female, 67% non-Hispanic Black, age M(SD) = 40.76 (12.42)) who completed baseline ratings of depressive symptoms and seven consecutive days of in-vivo data collection focused on cigarette smoking and associated mood and craving ratings.

Results.

Results indicated that depression moderates the prospective relationship between mood and smoking (even when controlling for craving), such that participants with higher levels of depressive symptoms smoke more cigarettes in response to an improvement in mood (relative to their average mood), whereas participants with lower levels of depressive symptoms smoke more in response to worsening mood states (relative to their average mood).

Conclusions.

Attempting to maintain better mood may be a motivating factor for smoking among depressed individuals. These findings may be helpful in tailoring smoking cessation treatment programs for people exhibiting depressive symptoms.

There is a significant association between cigarette smoking and depression. Negative affect (NA), defined as subjective distress that subsumes various aversive mood states (anger, contempt, disgust, guilt, fear, and nervousness) (Watson, Clark, & Tellegen, 1988), is an important feature of depression (American Psychiatric Association, 2013) and is often implicated as an antecedent to smoking (e.g., Baker, Piper, McCarthy, Majeskie, & Fiore, 2004). Indeed, numerous studies demonstrate strong associations between depression and smoking (see meta-analyses: Heckman et al., 2016; Luger, Suls, & Vander Weg, 2014) and individuals with a history of major depressive disorder (MDD) are more likely to smoke compared to individuals without a history of MDD (e.g., Breslau, Kilbey, & Andreski, 1991). This pattern is seen among individuals with sub-clinical levels of depression as well (Brandon, 1994).

Although depression is characterized by a variety of affective, cognitive, behavioral and somatic components (Buckner et al., 2015), the affective features of depression, including negative mood, appear to be uniquely involved in the depression-smoking relationship (Haas et al., 2004). It follows that smokers with depression are likely to use cigarettes to enhance their mood. However, the existing literature employing ecological momentary assessment (EMA) methodology, a type of intensive longitudinal study design focused on data collection in real time and in the natural environment (Shiffman, Stone & Hufford, 2008), provides incongruous results, with most studies demonstrating no relationship between mood and smoking (Rathbun, Shiffman, & Gwaltney, 2006; Shiffman & Rathbun, 2011; Shiffman & Paty, 2006; Shiffman et al., 2002).

Researchers have yet to fully account for these unexpected results, but one possibility is that an overall lack of association between mood and smoking masks a moderator effect such that depressed (but not non-depressed) smokers respond to mood fluctuations with increased smoking. For example, Fucito and Juliano (2009) conducted a mood induction study with 121 smokers (38% with depressive symptoms) and found that depressive symptoms moderate the relationship between mood and smoking, such that decreases in positive mood increased smoking more among depression prone smokers compared to smokers with fewer depressive symptoms. Similarly, Dahne and colleagues (2017) conducted a study with 73 young adult smokers demonstrating that NA moderates the relationship between depression and smoking, such that cigarettes become more valuable for individuals with elevated depressive symptoms. The current experimental literature thus provides some evidence to support the notion that depressed people are more motivated to smoke in response to changes in mood compared to nondepressed individuals, but additional research is needed especially considering researchers have yet to investigate the impact of depression on the relationship between mood and smoking using EMA methodology.

The current study examined how between-subject levels of depression predict the within-subject relationship between mood and smoking. We hypothesized that the mood-smoking relationship would be stronger for individuals with higher levels of depression. Additionally, we examined the impact of craving on this relationship, as considerable research demonstrates a relationship between craving and smoking (e.g., Wray, Gass, & Tiffany, 2013) and many theories of drug dependence implicate craving as a motivating factor in drug use (Drummond, 2001). It is our hope that gaining a more complete understanding of the complex relationship between depression, mood, and smoking will allow for the development of more effective interventions.

Method

Study Design and Procedures

Participants were 96 adult daily cigarette smokers. Data from this subsample were drawn from alarger longitudinal study called the “MOMENT Study” (Mixed Method E-cigareTte Study). Complete study protocol details are presented elsewhere (Pearson et al., 2016). Briefly, the “MOMENT Study” was an intensive longitudinal study that employed a mixed-methods design to examine the electronic cigarette (e-cigarette) initiation process among adult smokers. Participants completed four in-person visits, followed by an online follow-up survey 30-days post-study completion. The current project focuses on the initial baseline visit and daily smoking and mood observations collected throughout the first study week, prior to e-cigarette initiation. At baseline, participants confirmed current smoking status with an exhaled CO test (≥8 ppm) and completed a questionnaire assessing sociodemographics, depression levels, and tobacco use history. Subsequently, a research assistant described the EMA procedures to participants. Data were collected between August 2014 and July 2016, and thee study was approved by Chesapeake IRB, Colombia, MD (Pro00008526) (see: www.advarra.com/).

Participants

One hundred and seventeen participants were recruited after completing an online screening survey promoted through public online postings, paid advertisements, and physical flyers. Eligible participants were English-speaking adults aged 18 years or older residing in the Washington, DC, metro area who smoked at least eight cigarettes per day for the past five years. Polytobacco users, defined as having smoked a little cigar/cigarillo, large cigar, or hookah or using smokeless tobacco more than five times in the last 30 days, were excluded to simplify tobacco use reports. Additional eligibility criteria included a) no electronic cigarette (e-cigarettes) use in the last 30 days; b) interest in trying e-cigarettes; c) daily cell phone use; and d) ownership of a smartphone (i.e., iPhone or Android) with an unlimited text message plan to eliminate the need for a study-specific device. A complete list of eligibility criteria is available elsewhere (Pearson el al., 2016). All criteria were assessed by self-report. Eligible participants received up to $285 for their participation in the larger study.

One participant withdrew before completing their baseline visit. Two participants did not provide sufficient data to accurately assess for levels of depression. Eighteen participants’ (10 male) data were discarded as they did not complete a preestablished number of daily assessments (i.e., fewer than four random-prompt assessments per day for more than two days in a row during week 1 and/or reporting no cigarette use). These participants’ levels of depression did not significantly differ from the included participants’ scores (t=−1.77, p =.09). Our final sample, therefore, included 96 (51 women, 45 men) adult (≥18 years of age) smokers with an average age of 40.76 years (SD = 12.42). The majorityidentified as non-Hispanic Black (67%), with 26% non-Hispanic White4% Hispanic, and 3% identified as multi-racial or “other” race/ethnicity.

EMA Data Collection

EMA data collection during week 1 focused on cigarette use and associated mood and craving ratings, both assessed as single items to reduce time demands. Responses were collected via text message using participants’ personal cell phones and were time and date stamped. The EMA data collection system returned an error message to participants if they skipped items or entered out-of-range values, to reduce missing data.

Measures

Initial Measures

Center for Epidemiologic Studies Depression Scale Revised (CESD-10).

The CESD-10 (Andersen, Malmgren, Carter, & Patrick, 1994) is a short form of the Center for Epidemiologic Studies Depression Scale (CES-D) (Radloff, 1977), a self-report scale which assesses depression in community samples. The CESD-10, which presents ten of the 20 original CESD items, was developed to reduce response burden (Amtmann et al., 2014). Scores can range from 0 to 30 with higher scores reflecting more depressive symptoms. When compared to the full CESD (cutoff ≥ 16), the CESD-10 (cutoff score ≥10) demonstrates high predictive accuracy for classifying persons as having “depressive symptoms” (κ = .97) (Andersen et al., 1994). In the current sample, reliability was acceptable (α= .78).

Fagerström Test for Nicotine Dependence (FTND).

The FTND is a six item self-report questionnaire used to measure nicotine dependence (Heatherton, Kozlowski, Frecker, & Fagerstrom, 1991). Scores can range from 0 to 10, with higher scores indicating higher levels of nicotine dependence.

EMA Measures.

EMA data collection focused on cigarette use and associated mood and craving ratings. Responses were collected via text message using participants’ personal cell phones, eliminating the need to carry a study-specific device and the assessments were designed to be low burden to facilitate data given the high frequency of assessment periods.

Mood.

During the random prompt assessments, participants were asked “How is your mood right now?” and used a 10-point Likert scale to rate their mood from 1 (“extremely bad”) to 10 (“extremely good”).

Craving.

During the random prompt assessments, participants were asked “Right now, how much do you want to smoke a cigarette” and used a 10-point Likert scale to rate their level of craving from 0 (“I do not want a cigarette at all”) to 9 (“I very much want a cigarette”).

Smoking.

Participants were instructed to report all cigarette use by entering the number of cigarettes they smoked via the participant-initiated texting system.

Overview of analyses

All analyses were conducted using SPSS 25.0 and HLM for Windows Version 7.03. Due to the nested nature of the data we used hierarchical linear modeling (HLM), which employs maximum likelihood estimation and flexibly handles missing data (Bryk & Raudenbush, 1992). We created a two-level model in which daily observations (level 1) were nested within-person (level 2). At level 1, each participant’s repeated daily observations were used to generate a unique regression equation modeling the participant’s own relationship between mood and smoking. For all analyses, predictor variables and covariates in level 1 were group mean centered. In other words, the HLM program subtracted each participant’s mean mood score from each of their momentary observations of mood. The level 1 regression equation demonstrating the mood-smoking relationship is displayed below:

$$CI{G}_{ti}={\pi }_{\mathit{0}i}+{\pi }_{\mathit{1}i}*\left(MOO{D}_{ti}\right)+e_{ti}$$

where CIG_ti is participant i’s smoking (number of cigarettes smoked) at time t. The intercept (π_0i) is participant i’s smoking at their average mood. The slope (π_1i) is the change in participant i’s smoking for every one-unit increase in mood, with a steeper slope indicating a stronger relationship between mood and smoking. e_ti is the error term for person i at time t. At level 2, we estimated the effect of depression (measured by CESD-10 at baseline) on the within-subject relationship between mood and smoking. We controlled for gender, as prior research suggests that negative affect is more strongly correlated with smoking in women (McKee, Maciejewski, Falba, and Mazure, 2003), and depression in women is more commonly comorbid with smoking compared to men (Husky, Mazure, Paliwal, and McKee, 2008). Additionally, gender was the only sociodemographic variable associated with either the dependent or independent variables in this dataset, with men smoking more cigarettes per day on average, compared to women (t(94)=−2.02, p=.04). However, gender was unrelated to depressive symptoms (t(94)=.05, p=.97).

All predictor variables and covariates in level 2 were grand-mean centered, that is, the HLM program subtracted each participant’s mean depression score from the sample mean. The level 2 equations are displayed below:

$${\pi }_{\mathit{0}i}={\beta }_{\mathit{00}}+{\beta }_{\mathit{01}}*\left(CES{D}_i\right)+{\beta }_{\mathit{02}}*\left({\mathit{\text{GENDER}}}_i\right)+r_{\mathit{0}i}$$

$${\pi }_{\mathit{1}i}={\beta }_{\mathit{10}}+{\beta }_{\mathit{11}}*\left(CES{D}_i\right)+{\beta }_{\mathit{12}}*\left({\mathit{\text{GENDER}}}_i\right)+r_{\mathit{1}i}$$

We conducted several sets of analyses to explore the hypothesized moderating effect of depression on the relationship between mood and smoking: a) concurrent analyses in which mood was used to predict smoking at the same assessment period (Model 1: t); b) retrospective analyses in which smoking was used to predict mood at the next assessment period (Model 2: t−1); and c) prospective analyses in which mood was used to predict smoking at the next assessment period (Model 3: t+1).

Missing Data

Assessment periods at level 1 with missing data were deleted pairwise, such that discrete incomplete observations were excluded from analysis but participants with missing data at level 1 were not excluded from the overall multi-level model. However, HLM does not allow for missing data at level 2; accordingly, each participant must have reported their gender and provided data for a complete CESD-10 score to be included in the model. As noted, two participants did not provide enough data to compute a CESD-10 total score and were excluded from all analyses.

Results

Descriptive Analyses

See Table 1 for participant sociodemographic and tobacco use history across the entire sample and by level of depressive symptoms. The average CESD-10 score in our sample (M= 9.22) is similar to what has been noted in other non-psychiatric samples (Amtmann et al., 2014; Bun, Yuet, & Paul, 2007; Miller et al., 2008). Using a cutoff score of 10 or greater on the CESD-10 (Andersen et al., 1994), 36 participants (38%) in the current sample were considered to have “depressive symptoms”. The aggregated means and standard deviations of all outcome and level 1 predictor variables were calculated in HLM and are listed in Table 2.

Table 1.

Baseline Participant Sociodemographics and Tobacco Use History

	Total (N=96)		Low (n=60)		High (n=36)
	n/mean	%/SD	n/mean	%/SD	n/mean	%/SD
Gender
Male	45	47	28	48	17	46
Female	51	53	31	52	31	54
Race/ethnicity
Non-Hispanic Black	64	67	42	71	22	60
Non-Hispanic White	25	26	14	24	11	30
Muli-racial/other	3	3	3	5	4	10
Hispanic	4	4	3	5	1	3
Age	40.76	12.42	42.12	13.68	38.59	9.88
CES-D-10	9.22	4.02	6.73	1.71	13.19	3.40
Education
High school or less	30	31	19	54	11	30
Some college or more	66	69	40	46	26	70
Employment
Full-time or part-time	47	49	28	53	19	51
Not employed	49	51	31	47	18	49
Tobacco Use
Menthol preference	71	74	45	75	26	72
Participants making ≥1 past year quit attempt(s)	39	41	22	37	17	47
Age of first cigarette	16.28	5.60	16.00	4.16	16.75	7.44
Total years smoking	19.31	11.16	19.93	11.96	17.69	10.07
FTND	4.06	1.80	4.03	1.93	4.11	1.58

Table 2.

Means and Standard Deviations of Aggregated Predictor and Outcome Variables (N=96)

	M	SD
Mood	6.98	2.37
Craving	4.70	3.20
Cigarettes per assessment period	1.29	1.95

EMA Data

Over the course of seven days, participants self-initiated a total of 4,237 complete cigarette reports over the study week with an average of 6.06 (SD=0.31) reports per day. An additional 180 self-initiated reports were initiated but not completed. On average, participants reported smoking 7.83 cigarettes per day (SD=4.89).

Additionally, participants completed a total of 2,555 random prompts, or 62.39% of all random prompts, with the remaining 1,540 prompts incomplete. The average number or random prompts completed was 27.55 (SD=6.89). Compliance was approximately as expected of participants (i.e., completing 4 out of 6 random prompts per day). Random prompt completion did not vary by study day χ²(6)=8.07 p = .23 and was unrelated to level of depressive symptoms, F(1,94)=0.55, p=.46, or the number of cigarettes smoked per day, F(1,94)=1.00, p=.32.

Moderating Effect of Depression on the Mood-Smoking Relationship

See Table 3 for full details.

Table 3.

Multilevel Regressions: Effect of Depression on the Relationship Between Mood and Smoking (N=96)

	B Coefficient	SE	T Ratio
Model 1 (t)
Average smoking, β00	1.34***	0.08	15.96
Effect of depression on smoking, β01	0.02	0.02	0.88
Average mood-smoking slope, β10	0.02	0.02	1.21
Effect of depression on mood-smoking slope, β11	−0.00	0.00	−0.70
Model 2 (t−1)
Average smoking, β00	1.39***	0.08	16.83
Effect of depression on smoking, β01	0.02	0.02	1.12
Average mood-smoking slope, β10	0.03	0.02	1.61
Effect of depression mood-smoking slope, β11	0.01	0.00	1.86
Model 3 (t+1)
Average smoking, β00	1.31***	0.09	15.33
Effect of depression on smoking, β01	0.02	0.02	1.08
Average mood-smoking slope, β10	−0.01	0.02	−0.46
Effect of depression on mood-smoking slope, β11	0.01*	0.00	2.02

Note: Coefficients are unstandardized, from HLM output.

^*p<.05;

^**p<.01;

^***p<.001. df = 93.

Concurrent relationship.

The within-person intercept (β₀₀) represents the average number of cigarettes smoked per assessment period at on an average mood day with average levels of depression. The within-person intercept (β₀₀) did not vary by level of depression. The average within-person slope (β₁₀) (i.e. the average relationship between mood (t) and smoking (t) at average levels of depression) was non-significant, demonstrating no concurrent relationship between mood and smoking. There were no interactive effects of depression on the mood-smoking slope (β₁₁).

Retrospective relationship.

The average within-person slope (β₁₀) was non-significant, demonstrating no retrospective relationship between mood and smoking, that is, smoking did not predict mood at the next assessment period. There were no interactive effects of depression on the mood-smoking slope.

Prospective relationship.

The average within-person slope (β₁₀) was non-significant, demonstrating no prospective relationship between mood and smoking. However, the data reveal a significant interaction in which depression moderates the prospective relationship between mood and smoking (β₁₁ = 0.01, p = .04). Individuals higher in depression (≥1 standard deviation above the mean) had a positive within-person association between mood and subsequent cigarette smoking and individuals with lower levels of depression (≥1 standard deviation below the mean) had a negative within-person association between mood and subsequent cigarette smoking. Participants with high levels of depressive symptoms smoke more cigarettes in response to an improvement in mood relative to their average mood, whereas participants with lower levels of depressive symptoms smoke fewer cigarettes in response to an improvement in mood relative to their average mood (Figure 1).

Figure 1.

Prospective Mood-Smoking Relationship as a Function of Depression.

Additional analyses

Many theories of drug dependence implicate craving as a motivating factor in drug use (Drummond, 2001) and considerable research demonstrates a relationship between craving and smoking (e.g., Berkman, Falk, & Lieberman, 2011). Considering the association between craving and smoking, we added craving as a predictor variable at level 1 in Models 1–3 to create a new set of analyses (Models 4–6). Level 2 remained the same. See Table 4 for full details. All results were similar. Additionally, the concurrent relationship between craving and smoking when controlling for mood was significant (β₂₁ = 0.04, p < .01), such that the number of cigarettes smoked increases as craving increases in relation to one’s average craving rating. Similarly, there was significant retrospective relationship between craving and smoking when controlling for mood (β₂₁ = −0.048, p < .001), in which craving ratings decreased in relation to one’s average craving rating after smoking. However, depressive symptoms did not moderate the relationship between craving and smoking.

Table 4.

Multilevel Regressions: Effect of Depression on the Relationship Between Mood, Craving, and Smoking (N=96)

	B Coefficient	SE	T Ratio
Model 4 (t)
Average smoking, β00	1.35***	0.09	15.82
Effect of depression on smoking, β01	0.02	0.02	0.89
Average mood-smoking slope, β10	0.03	0.02	1.58
Effect of depression on mood-smoking slope, β11	−0.00	0.00	−0.83
Average craving-smoking slope, β20	0.04***	0.01	3.71
Effect of depression on craving-smoking slope, β21	−0.00	0.00	−0.40\
Model 5 (t−1)
Average smoking, β00	1.39***	0.08	16.82
Effect of depression on smoking, β01	0.02	0.02	1.08
Average mood-smoking slope, β10	0.02	0.02	1.05
Effect of depression mood-smoking slope, β11	0.01	0.00	1.53
Average craving-smoking slope, β20	−0.05***	0.01	−4.86
Effect of depression on craving-smoking slope, β21	−0.00	0.00	−0.75
Model 6 (t+1)
Average smoking, β00	1.31***	0.09	15.32
Effect of depression on smoking, β01	0.02	0.02	1.08
Average mood-smoking slope, β10	−0.01	0.02	−0.52
Effect of depression on mood-smoking slope, β11	0.01*	0.00	2.28
Average craving-smoking slope, β20	0.00	0.01	0.09
Effect of depression on craving-smoking slope, β21	0.00	0.00	0.85

Note: Coefficients are unstandardized, from HLM output.

^*p<.05;

^**p<.01;

^***p<.001. df=93

Discussion

Previous EMA studies of the association between mood and smoking typically showed null results. Our study extended this work by testing the moderating role of depressive symptoms on the mood-smoking relationship as smokers have an increased risk of depression compared to nonsmokers (Pasco et al., 2008). We also examined the role of craving in the mood-smoking relationship. Across all analyses, mood was unrelated to smoking: a) there was no reliable association between mood and concurrent smoking; b) mood did not significantly change in response to smoking at the previous assessment period; and c) mood did not predict smoking at the next assessment period. However, craving was associated with smoking. An increase in craving predicted a concurrent increase in cigarette smoking and there was a significant reduction in craving following a smoking event. Together, these results replicate the majority of the existing EMA literature on antecedents of smoking behaviors, which suggest that mood is unrelated to smoking (e.g., Carter et al., 2008), but that craving is a strong predictor of smoking behaviors (e.g., Dunbar, Scharf, Kirchner, & Shiffman, 2010).

Several explanations have been put forth in attempt to explain the null relationship between mood and smoking. First, it is possible that smokers may be initially motivated to smoke in response to mood changes, but that these associations fade with long-term cigarette use as smoking becomes driven by the need to regulate nicotine levels (Baker et al., 2004; Shiffman et al., 2002; Shiffman & Paty, 2006). Or, mood fluctuations may initiate a series of responses that eventually results in smoking, but by the time the actual smoking occurs, the affective antecedents that were present may no longer be salient to the participant (Shiffman et al., 2004).

The overall lack of an effect for mood on smoking may also reflect individual variance in depressive symptoms. Collapsing smokers with varying levels of depressive symptoms may have concealed existing associations between mood and smoking. Indeed, the current study results demonstrate that smokers with high levels of depression smoke more in response to an improvement in mood, whereas smokers with low levels of depression smoke more in response to worsening mood states. We hypothesized that depressive symptoms would moderate the relationship between mood and smoking, such that participants would smoke more cigarettes in response to a negative mood state and that this relationship would be stronger among those with higher levels of depressive symptoms. However, participants with high levels of depressive symptoms smoked more cigarettes in response to an improvement in mood relative to their average mood.

In considering the implications of these results, it is important to acknowledge that the assessment of affect is challenging considering disagreement regarding the operationalization of affective states and the long-standing debate regarding the independence (Watson et al., 1988) or bipolarity (Russell, 1980) of positive affect (PA) and negative affect (NA). We viewed affect as single dimension and participants were asked to describe their mood on one Likert scale ranging from “extremely bad” to “extremely good”. While this approach made the assessments easy and quick for participants to complete, results only allowed for an interpretation of whether mood improved or worsened. We do not have enough information to determine if improvement was based on a reduction in NA, an increase in PA, or both and vice versa for a worsening mood state.

As such, more research is needed to (a) replicate our results and (b) understand more precisely how fluctuations in PA and NA affect smoking behaviors among more vs. less depressed individuals. Until that research is completed, interpretation is speculative, but we believe it is worth considering the possibility that PA and NA might even be differentially relevant for more vs. less depressed smokers. That is, low-depression smokers, who smoked more following unfavorable mood, might be reacting to elevated NA by smoking. Regardless of its true value, the belief that smoking, for instance, calms anxious feelings is widespread and would make smoking more a seemingly logical response to elevated NA (Kassel, Stroud, & Paronis, 2003).

Conversely, we consider it plausible that the more depressed smokers might have used cigarettes to elevate low PA. Low PA, or anhedonia (the loss of interest in all, or most activities) is a core feature of MDD (APA, 2013), and may be a motivating factor for smoking among depressed individuals who have persistent difficulties experiencing PA in response to pleasurable events (Cook, Spring, & Mcchargue, 2007). A reduction in the ability to experience PA is believed to be linked with the attenuation of the mesolimbic system (reward center of the brain) (Rampello & Nicoletti, 2000), and by self-administering nicotine (a dopamine agonist) (Picciotto, Brunzell, & Caldarone, 2002) depressed smokers may be augmenting their capacity to experience pleasurable mood states (Cook et al., 2007).

There is growing evidence that anhedonia is a component of tobacco withdrawal syndrome (Cook et al., 2017), that abstinence from tobacco increases anhedonia (see systematic review: Hughes, Klemperer, & Peasley-Miklus, 2018), and that a history of anhedonia may make it more difficult to quit smoking (Leventhal et al., 2014).

Similarly, several studies demonstrate that nicotine administration has mood-enhancing effects (Stein et al., 1998; Warburton & Mancuso, 1998) and the association between smoking and PA is stronger among depressed individuals. For example, results from a study comparing depression-prone smokers (i.e., individuals with a current or history of MDD diagnosis) and smokers without a history of MDD demonstrate that depression-prone smokers have an enhanced positive mood response in reaction to a positive mood induction when smoking a nicotinized cigarette, compared to a denicotinized cigarette (Spring et al., 2008). These findings suggest that self-administering nicotine improves depression-prone smokers’ affective response to pleasant stimuli (Spring et al., 2008). Relatedly, a study of smokers with and without a history of MDD provides data demonstrating that smoking a nicotinized cigarette versus placebo during a positive mood induction allowed anhedonic smokers to experience a significant increase in PA, whereas nicotine did not significantly affect positive mood among less depressed smokers (Cook et al., 2007). Cook et al. (2007) posit that anhedonic/depression-prone individuals may learn to engage in compensatory behaviors, like smoking, to augment positive mood states.

Besides the unidimensional assessment of affect noted earlier, this study has additional methodological limitations that require cautious interpretation of the findings. First, the study utilized an online screener for initial participant recruitment and the protocol required participants to have an iPhone or Android smartphone, which may have been a barrier to enrollment for individuals lacking internet access and/or the requirement cellphone, respectively. However, approximately 72% individuals who were screened met all smartphone inclusion criteria, suggesting that smartphone ownership was not a major barrier to study participation for individuals with internet access (Harvey et al., 2018).

Despite the extensive benefits of EMA these methods also carry disadvantages. For example, EMA requires participants to respond to study prompts and report daily activities throughout the day, in this case for seven days. This intensive form of data collection is burdensome and subject to participant non-compliance, often resulting in missing or falsified data (Shiffman et al., 2008). To reduce non-compliance rates, study participants were compensated for their time and efforts and incentivized to complete as many assessments as possible with monetary bonuses. Additionally, a RA monitored participant adherence to EMA procedures and encouraged them to keep up with the daily assessments with emails or text messages. Participants responded to an average 61% of all random prompts, which was in line with our expectations (4 out of 6 random prompts per day; 66%) as it is unrealistic to expect no missing data when implementing intensive longitudinal methods (Bolger & Laurenceau, 2013).

At the same time, considering the pattern of missing data is important. Another study utilizing a different subset of the Moment Study data compared the accuracy of self-report data to objective data from a Bluetooth-enabled device and results suggested that the Bluetooth device captured significantly more data in relation to the standard EMA (Pearson et al., 2017). However, results suggest that participants underreported their smoking 90% of the time and while missing data is certainty not desired, a consistent pattern to missing data can be accounted for in data analysis and/or interpretation (Pearson et al., 2017). The consistent pattern of underreporting may explain the lower than expected average number of cigarettes smoked per day. Inclusion criteria required participants to smoke at least eight cigarettes per day, however the current EMA results demonstrate that participants smoked an average of 7.83 cigarettes per day (SD = 4.90).

Overall, the present findings demonstrate that depression moderates the prospective relationship between mood and smoking, such that individuals with higher levels of depression smoke more in response to an improvement in mood, whereas individuals with lower levels of depression smoke more in response to a worsening mood. These results provide further evidence that affect is associated with ad lib smoking but that the relationship is dependent on individual levels of depressive symptoms. Additionally, findings highlight the importance of studying depression within the context of tobacco research and offer implications for developing smoking cessation interventions to be more effective for individuals exhibiting depressive symptoms, especially considering that depressed smokers demonstrate greater difficulty with smoking cessation and maintaining abstinence compared to their non-depressed counterparts (Hitsman, Papandonatos, Mcchargue, Demott, & Spring, 2014). If smoking helps depressed smokers to extend favorable mood states, treatments for depressed smokers could benefit from an increase in focus on behavioral activation to minimize depressive symptoms (e.g., Behavioral Activation Treatment for Depression; Hopko, Lejuez, LePage, Hopko, & McNeil, (2003)) and discovering pleasurable activities to replace or serve as a competing-response behavior to smoking, as an improvement in mood may be a trigger for smoking among this group of individuals. To our knowledge, this is the first EMA study to examine the interactive effects of depression on the relationship between mood and smoking, thus additional research is needed to replicate the current findings and to understand the mechanisms contributing to the relationship.