Anxiety Sensitivity in Relation to Cigarette Smoking and Other Substance Use in African American Smokers

Abstract

Anxiety sensitivity (AS)—fearfulness of anxiety symptoms—has been implicated in the etiology of emotional disorders (e.g., depressive and anxiety disorders) and linked to cigarette smoking and other substance use (SU). However, studies examining AS in relation to SU primarily have been conducted with racially/ethnically heterogeneous or mostly European American samples. Hence, this cross-sectional study involving secondary analysis of baseline data focused on investigating associations of AS with cigarette smoking and other SU in a sample of 630 non-treatment-seeking African American smokers (37.3% female; M age = 49.6 years; M cigarettes smoked per day = 15.4). After screening out individuals with non-nicotine substance dependence, participants reported their demographics, AS, dysphoria symptoms (i.e., depression and anxiety symptoms), and SU. In regression analyses controlling for dysphoria symptoms, age, education level, income level, and years of regular smoking, AS was positively associated with tobacco withdrawal severity (β = .12, p = .007), overall smoking motives (β = .17, p < .001), alcohol use problems (β = .12, p = .005), and other (non-nicotine, non-alcohol) SU problems (β = .16, p < .001). Though lacking the passage of time between assessments needed to provide strong evidence of mediation, unplanned analyses further revealed indirect associations of AS with several SU variables through dysphoria symptoms. Current findings are consistent with those found in prior samples and suggest that AS is similarly related to SU in African Americans, who may benefit from interventions that have been helpful in improving AS, dysphoria symptoms, and SU in other groups.

Anxiety sensitivity (AS), a personality trait reflecting the tendency to fear anxiety symptoms and their potential negative consequences, has been meta-analytically linked to symptoms of several emotional disorders (e.g., panic, generalized anxiety, social anxiety, post-traumatic stress, and depressive disorders; Naragon-Gainey, 2010; Olatunji & Wolitzky-Taylor, 2009). Importantly though, AS has been shown to be conceptually and empirically distinct from specific emotional disorder symptoms (e.g., panic, general anxiety, and depression) as well as from negative emotion or negative affect (NA) in general (Cox, Borger, Taylor, Fuentes, & Ross, 1999; McNally, 2002; Naragon-Gainey, 2010; Qi, Rappaport, Cecilione, Hettema, & Roberson-Nay, 2019). Theoretically, individuals higher in AS (relative to individuals lower in AS) tend to have stronger beliefs that anxiety symptoms may result in unpleasant or dangerous outcomes (e.g., impending heart attack, worsening mental disturbance, others noticing anxiety), and consequently high-AS individuals tend to experience greater fear in response to anxiety symptoms, thereby contributing to the exacerbation of both momentary state anxiety in particular and state NA in general and less proximately (over time) to the development of dysphoria (i.e., depression and anxiety symptoms) and other emotional disorder symptoms (Naragon-Gainey, 2010; Olatunji & Wolitzky-Taylor, 2009; Reiss, Peterson, Gursky, & McNally, 1986). This theoretical model of AS and its relation to state NA and emotional disorder symptoms has been evidenced by associating AS with experimentally induced increases in acute physiological startle responses (Melzig, Michalowski, Holtz, & Hamm, 2008) and subjectively reported responses of state fear/panic, anxiety, and NA (Fluharty, Attwood, & Munafo, 2016; McNally, 2002). This theoretical model has been further evidenced by prospectively relating AS to the emergence and persistence of emotional disorder symptoms and diagnoses, including panic (McNally, 2002), generalized and social anxiety (Waszczuk, Zavos, & Eley, 2013), post-traumatic stress (Boffa et al., 2016), and depression (Qi et al., 2019). Moreover, two treatment studies have evidenced that the beneficial effects of treatment on depression and anxiety symptoms are mediated by reductions in AS (Norr, Allan, Macatee, Keough, & Schmidt, 2014; Smits et al., 2008). Hence, an abundance of research indicates that AS is involved in the etiology of emotional disorders.

Given that emotional disorders have been associated with substance use (SU) problems in numerous longitudinal studies (Kuo, Gardner, Kendler, & Prescott, 2006; O’Neil, Conner, & Kendall, 2011; Wolitzky-Taylor, Bobova, Zinbarg, Mineka, & Craske, 2012), it is possible that AS may also be involved in the development of SU. Indeed, a substantial and growing literature does evidence that AS is related to cigarette smoking and other SU. In particular, AS has been associated with both SU coping expectancies and motives (i.e., both expecting a substance to reduce NA if used and explicitly reporting the reason for using a substance is to reduce NA, respectively) in regard to cigarette smoking and alcohol and cannabis use (DeMartini & Carey, 2011; Guillot et al., 2019; Guillot, Blumenthal, Zvolensky, & Schmidt, 2018; Guillot, Leventhal, Raines, Zvolensky, & Schmidt, 2016). AS additionally has been associated with problems related to the use of various substances, including tobacco, alcohol, and cannabis (Guillot et al., 2019; Guillot et al., 2018). More specifically in regard to cigarette smokers, AS has been prospectively associated with more intense tobacco withdrawal and higher risk of smoking lapse and relapse (Bakhshaie et al., 2018). Based on prior work, it is plausible that some individuals higher in AS (relative to individuals lower in AS) may be more motivated to use substances to alleviate state NA (e.g., NA related to withdrawal or other stressors) and ultimately reduce emotional disorder symptoms (e.g., depression/anxiety or dysphoria symptoms), which thereby may negatively reinforce SU and increase risk for SU problems (Guillot et al., 2019; Guillot et al., 2016). Consistent with this notion, high-AS individuals (relative to low-AS individuals) have reported greater anxiety relief or affective stress dampening when they drink or smoke subsequent to a stressor (DeMartini & Carey, 2011; Guillot, Pang, & Leventhal, 2014).

Prior research on AS in relation to smoking or other SU has almost exclusively been conducted in racially/ethnically heterogeneous or mostly European American samples (summarized in the previous paragraph), with few such studies having focused on African American samples (Carter, Sbrocco, Lewis, & Friedman, 2001; Talkovsky & Norton, 2015). Similar to European American adult samples, however, AS has been positively correlated with symptom measures of panic, anxiety, depression, and general distress (or dysphoria) in samples of African American adults (Carter et al., 2001; Talkovsky & Norton, 2015). In addition, anxiety, depressive, and general negative affective symptoms have been concurrently and longitudinally related to smoking and other SU in African American samples (Bares & Andrade, 2012; Clark, 2014; Repetto, Caldwell, & Zimmerman, 2005), as in mixed race/ethnicity or mostly or entirely European American samples (Bares & Andrade, 2012; Marmorstein, White, Loeber, & Stouthamer-Loeber, 2010; Schuler, Vasilenko, & Lanza, 2015). Therefore, it is reasonable to suspect that AS would be related to smoking and other SU in African Americans similar to in other groups. To our knowledge, only three prior studies have investigated AS in relation to SU in samples that were largely composed of (i.e., consisted of ~75% or more) African Americans. In one cross-sectional study of cannabis users with a racial/ethnic minority background (nearly 75% African American), AS was associated with greater cannabis problems, hazardous drinking, and alcohol consumption and problems (Paulus, Manning, Hogan, & Zvolensky, 2017). In two cross-sectional studies of individuals in residential SU treatment (~85–95% African American), AS was positively related to heroin use (Lejuez, Paulson, Daughters, Bornovalova, & Zvolensky, 2006) and cigarettes smoked per day (Dahne, Hoffman, & MacPherson, 2015). Even less common, only two previous cross-sectional studies have examined AS in relation to SU in exclusively African American samples: The first study found that only the physical concerns aspect of AS (i.e., fear of somatic symptoms of anxiety) as measured by the Anxiety Sensitivity Index-3 (ASI-3) was related to greater cannabis use problems (Dean, Ecker, & Buckner, 2017). However, this study did not report if general AS was related to cannabis use problems, and though there is some initial evidence of the factor validity of the ASI-3 Physical Concerns dimension across different racial/ethnic groups (Jardin et al., 2018), this AS facet finding should be interpreted with caution given that prior research using the original ASI found that the items comprising the Physical Concerns factor in samples of mostly European American adults (Rodriguez, Bruce, Pagano, Spencer, & Keller, 2004) were better represented as two separate factors related to fears of unsteadiness and fears of cardiovascular problems in samples of African American adults (Hunter, Keough, Timpano, & Schmidt, 2012). Lastly, in the second study, AS was not directly associated with hazardous drinking, which was the only form of SU examined (Haas, Forkus, Contractor, & Weiss, 2019). In summary, general AS thus far has not been directly associated with smoking or other SU in an entirely African American sample.

Therefore, the purpose of the current cross-sectional study was to examine AS in relation to smoking and other SU in African American smokers. Based on prior research in samples consisting exclusively or mostly of African Americans, we hypothesized that AS would be positively associated with cigarettes smoked per day (Dahne et al., 2015) and with levels of hazardous drinking, alcohol consumption and problems (Paulus, Manning, et al., 2017), and other SU problems (Lejuez et al., 2006; Paulus, Manning, et al., 2017). Also, based on prior research in racially/ethnically heterogeneous or predominately European American samples, we hypothesized that AS would be positively associated with self-reported difficulty attempting to quit smoking (Zvolensky, Farris, Schmidt, & Smits, 2014), severity of tobacco withdrawal symptoms (Guillot, Zvolensky, & Leventhal, 2015), overall smoking motives (McLeish, Farris, Johnson, Bernstein, & Zvolensky, 2015), and barriers to smoking cessation (Zvolensky et al., 2014), but would not be related to tobacco dependence severity (Guillot, Zvolensky, et al., 2015).

Method

Participants

Participants were 630 African American daily cigarette smokers (37.3% female; M age = 49.6 years, SD = 11.0) recruited from the Los Angeles metropolitan area through paper and online advertisements and word of mouth. The current report reflects a secondary analysis of baseline (cross-sectional) data from a larger study focused on investigating individual differences in tobacco withdrawal in African American smokers (Bello, Pang, Chasson, Ray, & Leventhal, 2017; Liautaud, Leventhal, & Pang, 2018). To be included, participants had to self-identify as non-Latinx African American and be 18+ years of age and a smoker for 2+ years who typically smokes 10+ cigarettes per day. Exclusion criteria were being diagnosed with DSM-IV non-nicotine substance dependence during the baseline session; providing breath samples at intake indicative of a blood alcohol content (BAC) > .000 or a carbon monoxide (CO) level < 10 ppm (SRNT Subcommittee on Biochemical Verification, 2002); expressing a desire to reduce or quit smoking in the next 30 days; and self-reporting psychiatric medication use, pregnancy or breastfeeding, or daily non-cigarette tobacco or nicotine replacement use.

Due to not all participants having ever attempted to quit smoking, data were available for 511 and 502 participants, respectively, in regard to the Smoking History Questionnaire (SHQ) Quit Difficulty item and SHQ Withdrawal Severity scale. Also, due to some measures being added midway into the study, data were available for 500 and 464 participants, respectively, in regard to the Brief Wisconsin Inventory of Smoking Dependence Motives (WISDM-Brief) and Barriers to Cessation Scale (BCS). Data for other measures were available for the total sample of African American smokers (N = 630). Participants were compensated $200, and the University of Southern California IRB reviewed and approved study procedures (protocol #HS-13–00225).

Procedure

Following a telephone screen to determine preliminary eligibility, participants attended an intake session involving in-person screening and assessments. This initial session involved completing informed consent, breath CO and BAC analysis, the Structured Clinical Interview for DSM-IV Axis I Disorders, Research Version, Non-Patient Edition SU disorder module only (First, Spitzer, Gibbon, & Williams, 2002), and the self-report measures listed below.

Measures

Anxiety Sensitivity Index (ASI).

The 16-item ASI (Reiss et al., 1986) assesses the degree to which an individual is fearful of anxiety symptoms and their perceived potential negative consequences, including physical, mental, and social concerns. Participants rated items on a 5-point Likert scale ranging from 0 (very little) to 4 (very much), with higher total sum scores indicating higher levels of AS. The convergent, concurrent, and predictive validity of the ASI has been evidenced across numerous samples (McNally, 2002; Naragon-Gainey, 2010; Olatunji & Wolitzky-Taylor, 2009), including samples of African Americans (Carter et al., 2001; Talkovsky & Norton, 2015). Furthermore, the discriminant and incremental validity of the ASI has been established (Cox et al., 1999; McNally, 2002; Naragon-Gainey, 2010; Qi et al., 2019) by showing that AS is distinct from trait negative emotionality/neuroticism, anxiety, and depression, that AS relations with individual emotional disorders are uniformly strong while at the same time significantly different in magnitude, and that AS prospectively predicts panic/anxiety responses and panic, anxiety, and depressive symptoms incremental to that of baseline emotional disorder symptoms. Because the ASI subfactors have differed between African American and (predominately) European American samples (evidencing 3–4 subscales; Hunter et al., 2012; Rodriguez et al., 2004), as well as differed between the only two African American samples in which factor analyses were conducted (Carter, Miller, Sbrocco, Suchday, & Lewis, 1999; Talkovsky & Norton, 2015), in the current study we chose to stick to using only the ASI total score (general AS), which served as the principal predictor in primary analyses.

Inventory of Depression and Anxiety Symptoms (IDAS) Dysphoria Scale.

The IDAS (Watson et al., 2007) Dysphoria scale is a 10-item measure that assesses general distress, or more specifically, depression and anxiety symptoms common to emotional disorders. Participants rated how much they experienced each symptom during the past two weeks on a scale ranging from 1 (Not at all) to 5 (Extremely). Thus, higher sum scores on the IDAS Dysphoria scale indicate greater dysphoria symptoms. The IDAS has been shown to have strong convergent and discriminant validity (Watson et al., 2007), with IDAS Dysphoria scores in particular correlating positively with other measures of depression and anxiety symptoms (r = .64-.84) and with measures of panic (r = .63), social anxiety (r = .66), and PTSD symptoms (r = .57). The IDAS Dysphoria score was used as a planned covariate and secondary predictor in primary analyses.

Demographic Questionnaire.

An author-constructed questionnaire assessed age, gender, education, relationship status, and income, which are reported as sample characteristics and were tested in preliminary analyses as potential unplanned covariates for primary analyses.

Alcohol Use Disorder Identification Test (AUDIT).

The 10-item AUDIT (Babor, Higgins-Biddle, Saunders, & Monteiro, 2001) assesses drinking patterns and consequences related to drinking. Participants respond to items on a scale of 0–4, where 0 indicates minimal or lack of occurrence and 4 indicates the presence of or a high frequency of occurrence. The AUDIT yields a total sum score for hazardous drinking and has been factor analyzed into two subscales: Alcohol Consumption (i.e., frequency/heaviness of consumption) and Alcohol Problems (i.e., frequency of alcohol-related consequences). Since the two-factor structure and convergent/concurrent validity of the AUDIT have been evidenced across a wide diversity of samples (Peng, Wilsnack, Kristjanson, Benson, & Wilsnack, 2012), we included the AUDIT total score (hazardous drinking) and its two subfactors (alcohol consumption and problems).

Drug Abuse Screening Test (DAST-10).

The DAST-10 (Cocco & Carey, 1998) consists of 10 items and assesses non-nicotine, non-alcohol SU and associated problems during the past year. Though many examples of other drugs are provided, the DAST-10 does not ask about the use of specific drugs and thus pertains to other SU in general. Respondents are asked to circle “Yes” or “No” for each item, with a “Yes” response scored as 1 and a “No” response scored as 0, except for of Item #3 which is reverse-scored. Item scores are summed to produce a total score of 0–10, with higher total scores representing a higher level of other SU problems. The DAST-10 has evidenced good convergent and concurrent validity (Cocco & Carey, 1998).

Fagerström Test for Cigarette Dependence (FTCD).

The FTCD (Fagerström, 2012; Heatherton, Kozlowski, Frecker, & Fagerström, 1991) is a 6-item measure that assesses tobacco dependence severity specific to cigarettes, with greater sum scores (range: 0–10) indicating more severe tobacco dependence. The FTCD has been shown to have good convergent, concurrent, and predictive validity (Meneses-Gaya, Zuardi, Loureiro, & Crippa, 2009; Piper et al., 2008).

Smoking History Questionnaire (SHQ).

The SHQ is an author-constructed questionnaire that assesses usual number of cigarettes smoked per day (SHQ Cigs/Day item), difficulty with the most recent attempt to quit smoking (SHQ Quit Difficulty item), and severity of tobacco withdrawal experienced during the most recent quit attempt (SHQ Withdrawal Severity scale). SHQ Quit Difficulty item responses were dichotomized, with 1 representing that the most recent attempt to quit smoking was Easy or Slightly Difficult and 2 representing that the most recent attempt to quit was Difficult or Very Difficult. The 11-item SHQ Withdrawal Severity scale pertains to tobacco withdrawal symptoms evidenced in prior research (Guillot, Stone, et al., 2015; Hughes, 2007; Leventhal, Waters, Moolchan, Heishman, & Pickworth, 2010), including craving, irritability, nervousness, depression/sadness, difficulty concentrating, physical symptoms, difficulty sleeping, restlessness, increased eating, weight gain, and loss of interest or pleasure. Participants rated items on a 7-point Likert scale ranging from 1 (not true of me at all) to 7 (very true of me), with higher sum scores indicating more severe tobacco withdrawal. The SHQ Withdrawal Severity scale evidenced good convergent validity by correlating with scores on the FTCD (r = .31), WISDM-Brief (r = .55), and BCS (r = .65) in the current study.

Brief Wisconsin Inventory of Smoking Dependence Motives (WISDM-Brief).

The WISDM-Brief (Smith et al., 2010) is 37-item questionnaire that examines motives for cigarette smoking related to tobacco dependence. Respondents are asked to rate their degree of agreement with each item on a Likert scale ranging from 1 (not true of me at all) to 7 (extremely true of me). In prior samples of African American and mostly European American smokers (Smith et al., 2010), the WISDM-Brief was factor analyzed into 11 smoking motive subscales related to affiliative attachment, automaticity, loss of control, cognitive enhancement, craving, cue exposure/associative processes, social/environmental goads, taste, tolerance, weight control, and affective enhancement. In order to limit the number of statistical tests, however, we decided that only overall smoking motives (WISDM-Brief total scores) would be included in the current study. The WISDM-Brief total score is computed by summing the mean score for each subscale, with higher total scores indicating greater overall smoking motives. The WISDM-Brief has evidenced good convergent, concurrent, and predictive validity in samples of African American and mostly European American smokers (Smith et al., 2010).

Barriers to Cessation Scale (BCS)

The BCS (Macnee & Talsma, 1995) is a 19-item measure that assesses perceived stressors associated with attempting to quit smoking. Each item is rated on a 4-point Likert scale ranging from 0 (not a barrier) to 3 (large barrier), with higher total BCS scores indicating greater barriers to smoking cessation. Though two previous studies factor analyzed the BCS into three subscales related to addiction, external, and internal barriers (Garey et al., 2017; Macnee & Talsma, 1995), both samples in which factor analyses were conducted were predominately European American. For this reason, we chose to use only the BCS total score in the current study. The BCS has demonstrated adequate convergent and predictive validity in prior research (Garey et al., 2017; Macnee & Talsma, 1995).

Data Analysis

For preliminary analyses, we first examined the bivariate correlations between demographic/smoking history variables and predictor variables (i.e., AS, the primary predictor, and dysphoria symptoms, the planned covariate and secondary predictor), as well as bivariate correlations between predictor variables and criterion (SU) variables, in order to determine if any unplanned covariates would be used in primary analyses (see Preliminary Analyses for a summary of statistically significant correlations). Then we obtained descriptive statistics, Cronbach’s alphas, and correlations between AS, dysphoria symptoms, and SU variables.

Primary analyses involved conducting a total of 10 separate linear or binomial logistic regression analyses with AS or dysphoria symptoms (run simultaneously in each regression model) as the predictor of each outcome (SU) variable controlling for age, education level, income level, and years of regular smoking. All 20 regression results are reported either as an odds ratio (OR) with a confidence interval (CI) or as a standardized regression coefficient (β). Alpha (two-tailed) was set at .05, which was unadjusted for preliminary analyses. To correct for multiple tests, primary analyses used the Holm-Bonferroni method (Holm, 1979) to adjust alpha for each of the 10 AS results and each of the 10 dysphoria results. Though we considered conducting a multivariate regression analysis that incorporated the predictor and outcome variables in a single model rather than conducting separate regression analyses, a multivariate regression analysis is not recommended unless all outcome variables are at least moderately correlated, which was not the case with current outcome variables (as shown in Table 1 in regard to the minimal correlations between several SU variables; Teixeira-Pinto, Siddique, Gibbons, & Normand, 2009; UCLA Institute for Digital Research and Education, 2020).

Table 1.

Means, Standard Deviations, Cronbach’s Alphas, and Bivariate Correlations between Anxiety Sensitivity, Dysphoria Symptoms, and Substance-Related Measures

	M (SD) or %	1	2	3	4	5	6	7	8	9	10	11	12
1 ASI	20.2 (13.5)	(.91)
2 IDAS Dysphoria	17.9 (7.7)	.41†	(.90)
3 SHQ Cigs/Day	15.4 (7.9)	.09*	.02	NA
4 SHQ Quit Difficulty	72.5	.13**	.12**	.11*	NA
5 SHQ Withdrawal Severity	30.2 (11.0)	.29†	.43†	12**	.31†	(.91)
6 FTCD	5.5 (2.0)	.14†	.13**	.47†	.28†	.31†	(.56)
7 BCS	20.9 (12.8)	.26†	.43†	.05	.30†	.65†	.34†	(.92)
8 WISDM-Brief	154.2 (51.9)	.28†	.32†	.19†	.32†	.55†	.53†	.65†	(.96)
9 AUDIT Total Scale	3.7 (5.0)	.16†	.22†	.02	.00	.06	−.01	.06	.06	(.84)
10 AUDIT EtOH Consumption	2.3 (2.3)	.06	.14†	−.02	−.07	−.02	−.10*	.01	−.01	.79†	(.72)
11 AUDIT EtOH Problems	1.5 (3.5)	.19†	.23†	.04	.05	.10*	.05	.08	.09*	.91†	.48†	(.85)
12 DAST-10	2.0 (2.2)	.26†	.25†	.08	.10*	.19†	.11*	.13**	.07	.31†	.15†	.35†	(.80)

Note. N = 630, except for 511 for SHQ Quit Difficulty and 502 for SHQ Withdrawal Severity due to not all participants having ever attempted to quit smoking, and 500 for WISDM-Brief and 464 for BCS due to the measures being added midstream into the study; ASI = Anxiety Sensitivity Index; IDAS = Inventory of Depression and Anxiety Symptoms; SHQ = Smoking History Questionnaire; FTCD = Fagerström Test for Cigarette Dependence; BCS = Barriers to Cessation Scale; WISDM-Brief = Brief Wisconsin Inventory of Smoking Dependence Motives; AUDIT = Alcohol Use Disorders Identification Test; DAST-10 = Drug Abuse Screening Test 10-item version. Cronbach’s alphas (if applicable) are in parentheses along the top diagonal.

^*p < .05,

^**p < .01,

^†p < .001

Results

Preliminary Analyses

At the preliminary p < .05, AS was significantly correlated with age (r = .09), education level (r = −.11), and years of regular smoking (r = .10); dysphoria symptoms were significantly correlated with income level (r = −.09). Further, age (r = .09), education level (r = −.14), income level (r = −.12), and years of regular smoking (r = .12) were significantly correlated with other (non-nicotine, non-alcohol) SU problems (DAST-10 scores). Thus, age, education level, income level, and years of regular smoking were used as unplanned covariates in primary analyses.

The large majority of participants were not married or living with a partner (82.5%), and 85.6% had received a high school diploma or equivalency. On average, participants began regularly smoking at 19.9 years of age (SD = 6.2) and had been regularly smoking for 29.7 years (SD = 12.4). Means, standard deviations, Cronbach’s alphas, and bivariate correlations between AS, dysphoria symptoms, and substance-related measures are reported in Table 1. Results indicated that each scale or subscale demonstrated adequate-to-high internal consistency (α = .72-.96) with the exception of the FTCD showing rather low internal consistency (α = .56), which is typical of this well-validated measure (Meneses-Gaya et al., 2009).

Primary Analyses

Individual regression models between AS or dysphoria symptoms and substance-related measures expressed as β (SE) or OR (CI) are shown in Table 2. In addition, the semi-partial correlation coefficient squared (sr²) is included in text as a measure of effect size (Fritz, Morris, & Richler, 2012) after each SU variable with which AS was significantly associated (see below).

Table 2.

Individual Regression Models between AS or Dysphoria Symptoms and Substance-Related Measures Expressed as β (SE) or OR (CI)

	ASI	IDAS Dysphoria
Measure
Cigarette Use, Problems, and Motives
SHQ Cigarettes Per Day	.07 (.04)	.00 (.05)
SHQ Quit Difficulty	1.27 (1.02–1.58)	1.24 (0.90–1.56)
SHQ Withdrawal Severity	.12 (.04)*	.38 (.04)**
FTCD	.08 (.04)	.10 (.04)
BCS	.10 (.05)	.41 (.05)**
WISDM-Brief	.17 (.05)**	.25 (.05)**
Alcohol Use and Problems
AUDIT (Hazardous Drinking)	.09 (.04)	.19 (.04)**
Factor 1: Alcohol Consumption	.02 (.04)	.14 (.04)*
Factor 2: Alcohol Problems	.12 (.04)*	.18 (.04)**
Other Substance Use Problems
DAST-10	.16 (.04)**	.18 (.04)**

Note. N = 630, except for 511 for SHQ Quit Difficulty and 506 for SHQ Withdrawal Severity due to not all participants having attempted to quit smoking, and 500 for WISDM-Brief and 464 for BCS due to the measures being added midway into the study; ASI = Anxiety Sensitivity Index; IDAS = Inventory of Depression and Anxiety Symptoms; SHQ = Smoking History Questionnaire; FTCD = Fagerström Test for Cigarette Dependence; BCS = Barriers to Cessation Scale; WISDM-Brief = Brief Wisconsin Inventory of Smoking Dependence Motives; AUDIT = Alcohol Use Disorders Identification Test; DAST-10 = Drug Abuse Screening Test 10-item version. Each individual regression model involved the ASI or IDAS Dysphoria as the predictor controlling for the other variable and age, education, income, and years of regular smoking.

^*p < .05,

^**p < .01 (Holm-Bonferroni corrected)

Controlling for dysphoria symptoms, age, education level, income level, and years of regular smoking, AS was significantly associated with greater tobacco withdrawal severity (sr² = .01), overall smoking motives (sr² = .02), alcohol problems (sr² = .01), and other SU problems (sr² = .02), with unadjusted p-values of p = .007, p < .001, p = .005, and p < .001, respectively, and Holm-Bonferroni corrected p-values of p = .049, p < .01, p = .04, and p < .01, respectively. AS was not significantly associated with cigarettes per day, difficulty attempting to quit smoking, tobacco dependence severity, barriers to smoking cessation, or alcohol consumption.

Controlling for AS, age, education level, income level, and years of regular smoking, dysphoria symptoms were significantly associated with greater severity of tobacco withdrawal (sr² = .12), barriers to smoking cessation (sr² = .11), overall smoking motives (sr² = .05), hazardous drinking (sr² = .03), alcohol consumption (sr² = .02), alcohol problems (sr² = .02), and other SU problems (sr² = .03), with an unadjusted p-value of p < .002 and a Holm-Bonferroni corrected p-value of p < .01 for each finding. Dysphoria symptoms were not significantly related to cigarettes per day, difficulty attempting to quit smoking, or tobacco dependence severity.

Unplanned Analyses

Lilienfeld (2017, p. 155) noted that adjusting analyses for a particular covariate may constitute statistical “overcontrol” if the covariate is influenced by the construct of interest. Because AS has been implicated in the development of emotional disorder symptoms (e.g., depression/anxiety or dysphoria symptoms) and emotional disorder symptoms are suspected to mediate (at least part of) the relationship between AS and SU (Guillot et al., 2019), we decided to conduct unplanned total and indirect effect analyses using PROCESS version 3.4 for SPSS (Hayes, 2018) with AS, dysphoria symptoms, and five selected SU variables: More specifically, the total effects of AS on SU variables and the indirect effects of AS on SU variables through dysphoria symptoms were tested for the four SU variables that AS was significantly related to in primary analyses (i.e., tobacco withdrawal severity, overall smoking motives, alcohol problems, and other SU problems) and for one other SU variable (i.e., barriers to smoking cessation) in which the value of the unadjusted (bivariate) correlation with AS in Table 1 was nearly three times as large as the value of the adjusted standardized regression coefficient associated with AS in Table 2. Indirect associations of AS through dysphoria symptoms were tested using 10,000 percentile bootstrap samples (Hayes, 2018) and 99% CIs to adjust for multiple tests. Each total effect is reported as a standardized regression coefficient (β), with R² reported as an effect size measure. Each indirect effect is also reported as a standardized regression coefficient (β), which doubles as an effect size measure, the completely standardized indirect effect (Cheung, 2009).

Controlling for age, education level, income level, and years of regular smoking, total effect analyses revealed that AS was related to greater tobacco withdrawal severity (β = .28, p < .001, R² = .11), barriers to smoking cessation (β = .27, p < .001, R² = .10), overall smoking motives (β = .27, p < .001, R² = .11), alcohol problems (β = .19, p < .001, R² = .05), and other SU problems (β = .23, p < .001, R² = .10). Also, AS was indirectly related to tobacco withdrawal severity (β [99% CI] = .16 [.10 - .22]), barriers to smoking cessation (β [99% CI] = .17 [.11 - .25]), overall smoking motives (β [99% CI] = .10 [.05 - .16]), alcohol problems (β [99% CI] = .07 [.02 - .14]), and other SU problems (β [99% CI] = .07 [.03 - .13]) through dysphoria symptoms.

Discussion

To our knowledge, this is the first study to examine AS in relation to cigarette smoking in an entirely African American sample and is also the first study to report associations of general AS with variables related to smoking, drinking, and other SU in African Americans. In primary (regression) analyses controlling for dysphoria symptoms, age, education level, income level, and years of regular smoking, AS was positively associated with tobacco withdrawal severity, overall smoking motives, alcohol problems, and other SU problems in the current cross-sectional sample of 630 non-treatment-seeking African American smokers who were not diagnosed with any form of non-nicotine substance dependence. All four statistically significant AS associations from primary analyses summarized above are consistent with expectations based on prior research in other groups. More specifically, as with prior research conducted in racially/ethnically heterogeneous or predominately European American samples, AS was positively associated with tobacco withdrawal severity (Guillot, Zvolensky, et al., 2015) and overall smoking motives (McLeish et al., 2015). Also, in accord with prior work conducted with largely African American samples, AS was positively related to alcohol and other SU problems (Lejuez et al., 2006; Paulus, Manning, et al., 2017).

Notably, prior research has associated AS both with SU coping expectancies and motives (i.e., expecting a substance to reduce NA if used and using a substance for the purpose of reducing NA) in regard to smoking, drinking, and cannabis use (DeMartini & Carey, 2011; Guillot et al., 2019), and previous studies in African American samples have associated AS with depression and anxiety symptoms (Carter et al., 2001; Talkovsky & Norton, 2015) and depression and anxiety symptoms with SU (Bares & Andrade, 2012; Clark, 2014; Repetto et al., 2005). Similarly, the current study associated AS with dysphoria (depression/anxiety) symptoms in preliminary correlation analyses and further associated AS and dysphoria symptoms with a number of SU variables in primary analyses. In addition, AS was indirectly related to several SU variables (e.g., alcohol and other SU problems) through dysphoria symptoms in unplanned analyses, which is consistent with previous cross-sectional studies that indirectly associated AS with SU problems through emotional disorder symptoms (Allan, Albanese, Norr, Zvolensky, & Schmidt, 2015; Guillot et al., 2019; Lechner et al., 2014), including in African American college students (Haas et al., 2019). Thus, when viewed altogether, it is possible that some African Americans higher in AS (relative to those lower in AS) may be more motivated to use substances in order to alleviate state NA (e.g., related to withdrawal or other stressors) and ultimately reduce emotional disorder symptoms (e.g., depression/anxiety or dysphoria symptoms), which thereby may negatively reinforce SU and increase risk for SU problems, as suggested by prior work in other groups (Guillot et al., 2019; Guillot et al., 2016; Guillot et al., 2014).

Contrary to hypotheses and some prior work, AS was not related to levels of hazardous drinking or alcohol consumption (Paulus, Manning, et al., 2017). However, the only previous study to examine AS in relation to alcohol use in an entirely African American sample (Haas et al., 2019) did not find that AS was directly related to hazardous drinking (i.e., AUDIT total scores), which was replicated in the current study. One possibility is that AS may be related more to alcohol problems and less to alcohol consumption, which further tends to weaken the relationship between AS and hazardous drinking since AUDIT total scores partially consist of AUDIT Alcohol Consumption (with the other AUDIT subfactor being Alcohol Problems). Indeed, though several studies have reported a relation between AS and greater alcohol consumption (Cox, Swinson, Shulman, Kuch, & Reichman, 1993; Lammers, Kuntsche, Engels, Wiers, & Kleinjan, 2013; Paulus, Manning, et al., 2017; Stewart, Peterson, & Pihl, 1995; Stewart, Zvolensky, & Eifert, 2001), the large majority of studies have not found evidence of an association between AS and greater alcohol consumption (Buckner, Bonn-Miller, Zvolensky, & Schmidt, 2007; Chavarria et al., 2015; Guillot et al., 2018; Johnson et al., 2008; Leyro, Zvolensky, Vujanovic, & Bernstein, 2008; Novak, Burgess, Clark, Zvolensky, & Brown, 2003; Paulus, Valadka, et al., 2017; Paulus, Vujanovic, & Wardle, 2016; Woicik, Stewart, Pihl, & Conrod, 2009; Zvolensky et al., 2007), including the current study. Also concordant with this possibility, the evidence for an association between AS and hazardous drinking (AUDIT total scores) has been mixed, with some studies reporting evidence of this association (Guillot et al., 2018; Paulus, Manning, et al., 2017; Zvolensky et al., 2019; Zvolensky et al., 2014) and other studies not finding evidence of this association (Bakhshaie, Zvolensky, Allan, Vujanovic, & Schmidt, 2015; Haas et al., 2019; Paulus, Valadka, et al., 2017; Zvolensky, Kotov, Antipova, & Schmidt, 2003), including the current study. In contrast, AS consistently has been related to greater alcohol problems (Chavarria et al., 2015; Guillot et al., 2019; Guillot et al., 2018; Lammers et al., 2013; Paulus, Manning, et al., 2017; Paulus et al., 2016; Schmidt, Buckner, & Keough, 2007; Woicik et al., 2009; Wolitzky-Taylor et al., 2015), with rare exception (Buckner et al., 2007). Still, research conducted with African American samples of differing characteristics is needed to determine if AS is related more to alcohol problems and less to alcohol consumption in African Americans as generally has been found in other groups.

Though the current nonsignificant relationship between AS and tobacco dependence severity was expected (Guillot, Zvolensky, et al., 2015), the lack of significant associations of AS with cigarettes smoked per day (Dahne et al., 2015), difficulty attempting to quit smoking (Zvolensky et al., 2014), and barriers to smoking cessation (Zvolensky et al., 2014) in primary analyses were all unexpected. In regard to difficulty attempting to quit smoking and barriers to smoking cessation, the unexpected null results may be related to sample differences since all prior AS research in relation to these two variables was conducted with racially/ethnically heterogeneous or predominately European American samples. It also is worth noting that the previous study which reported an association between AS and cigarettes per day in a largely African American sample (Dahne et al., 2015) only examined bivariate correlations separately for men and women, which yielded a significant correlation between AS and cigarettes per day specifically in women (r = .20). Of further note, the nonsignificant association between AS and cigarettes per day in the current sample of African Americans is consistent with nearly all prior research conducted in other samples (Guillot et al., 2016; Guillot et al., 2014; Johnson et al., 2008; Leyro et al., 2008; Zvolensky et al., 2003), which evidences that AS tends to be unrelated to smoking heaviness. Though non-significant in primary analyses, in the current study AS was associated with barriers to smoking cessation both in preliminary bivariate correlation analyses (r = .26) and in unplanned total effect analyses in which all covariates except for dysphoria symptoms were statistically controlled (β = .27). Of additional interest, the only two previous studies to examine AS and cigarette smoking in Latinx samples yielded correlations that appear to be substantially larger in magnitude relative to current and prior correlations: In one study conducted in Mexican smokers (Zvolensky et al., 2007), AS was correlated with cigarettes per day (r = .32), and in a study conducted in Latinx smokers (Zvolensky et al., 2019), AS was correlated with tobacco dependence severity (r = .41), smoking quit attempt problems (r = .55), and barriers to smoking cessation (r = .56). Because AS-smoking relations are so understudied beyond that of racially/ethnically heterogeneous or predominately European American samples, it is important that future research into AS associations with smoking be conducted in samples exclusively consisting of African Americans, Latinx individuals, and individuals from other minority groups (e.g., Asian Americans) and statistically control for relevant covariates.

In the current study, dysphoria symptoms were positively associated with tobacco withdrawal severity, overall smoking motives, barriers to smoking cessation, hazardous drinking, alcohol consumption and problems, and other SU problems in primary analyses. These dysphoria symptom associations followed a pattern similar to that of AS associations. Also, the current associations between dysphoria symptoms and SU variables are consistent with prior research in African Americans that associated depression and anxiety symptoms with tobacco, alcohol, and other SU (Bares & Andrade, 2012; Clark, 2014; Repetto et al., 2005). The current study expands on this literature to some extent, given that the cited prior research used only one or a few items to assess SU and did not include measures of SU problems.

The present study has several limitations. A major limitation is the use of cross-sectional data, which hinders the ability to infer the directionality and causality of relations. Future studies on AS and SU in African Americans using longitudinal and quasi-experimental designs may be more useful in this regard, though further research would be welcome in this understudied area even with additional cross-sectional studies. Another limitation is that the effect sizes of adjusted AS associations with SU variables from primary (regression) analyses were rather small. A point worth considering, however, is that the values from primary analyses represent the strength of relations of AS with SU variables beyond that shared with covariates, which included dysphoria symptoms. As explained earlier, dysphoria symptoms are theorized to mediate (at least part of) the relationship between AS and SU, which means that by controlling for dysphoria, AS possibly is being stripped of some of its more proximate (downstream) effects on SU. Consistent with this notion, unplanned total effect analyses (which statistically controlled for all covariates except for dysphoria symptoms) suggest that the effect sizes of AS-SU relations from primary analyses involved being substantially decreased by adjustment for dysphoria symptoms. To elaborate, effect sizes for AS-SU relations from primary analyses were uniformly small (Fritz et al., 2012), whereas effect sizes of AS-SU relations from total effect models generally were more medium (Cohen, 1992). Also, though limited by the lack of time lapse between measures (required for providing more than suggestive evidence of mediation), unplanned indirect effect analyses suggest that dysphoria symptoms may mediate part of the relationship between AS and SU, as evidenced by indirect associations of AS with several SU variables through dysphoria symptoms (small-to-medium in effect size; Cheung, 2009). In regard to practical significance, findings from the current sample indicate that the potential influence of AS on SU which is unique from dysphoria may tend to be small in magnitude, but the potential combined influence of AS on SU that may be exerted both directly (separate from dysphoria symptoms) and indirectly through dysphoria symptoms may tend to be more medium in magnitude and have greater clinical relevance; in other words, considering only the potential effect of AS on SU when statistically controlling for dysphoria symptoms may underestimate the overall importance of AS in the development of SU if AS affects SU in part through its influence on dysphoria symptoms.

Another limitation of the current study is that, except for the AUDIT, only total scores for each substance-related outcome were used. For example, the WISDM-Brief was previously factor analyzed into 11 subscales in African American and mostly European American samples of smokers (Smith et al., 2010), which could have provided information on specific smoking motives instead of just on overall smoking motives. However, we decided that the two-fold increase in the number of regression analyses that would have to be conducted due to including the subscales from this measure was not warranted for the purpose of this initial investigation. A further limitation is that the DAST-10 assesses any SU problems besides those related to nicotine and alcohol. Future studies investigating AS-SU associations in African Americans may be more informative if measures assessing other specific forms of SU (e.g., sedative, cannabis, and opioid use) are included. Another limitation is the use of the original ASI. Whether AS is viewed as a single broad factor, or as the three subfactors of AS physical, cognitive, and social concerns that often have been replicated, the ASI-3 has shown improved reliability and validity relative to the original ASI (Jardin et al., 2018; Rodriguez et al., 2004; Taylor et al., 2007). Lastly, though it is a strength that the current sample is entirely African American, it also is specific to non-treatment-seeking smokers who were not diagnosed with any form of non-nicotine substance dependence. Thus, one should be cautious in generalizing current findings to other groups because AS relations with SU variables may differ in clinical samples or even in nonclinical samples in which individuals with moderate-to-severe non-nicotine SU disorders have not been excluded.

Despite these limitations, this study expands on the literature by presenting novel findings on AS associations with cigarette smoking and other SU in African American smokers. This information might be used to better understand the difficulties that African Americans with high AS may face in regard to persistent smoking and concomitant problems related to alcohol and other SU. One potential implication of this study is that interventions which focus on decreasing AS may also be helpful in reducing dysphoria symptoms and SU in African Americans. Such AS-focused interventions may consist of psychoeducation, cognitive restructuring, and interoceptive exposure, which in predominately European American treatment samples, have evidenced reductions in AS accompanied by decreased anxiety and depression (Mitchell, Capron, Raines, & Schmidt, 2014; Norr et al., 2014) and improvements in smoking abstinence (Zvolensky et al., 2018) and alcohol use (Wolitzky-Taylor et al., 2018).