Temporal Sequence of Incident Cigarette, Coffee, and Alcohol Use Among AA Participants

FULL REVIEW

Temporal sequence of incident cigarette, coffee, and alcohol use among AA participants

Michael S. Reich, M.D., Mary S. Dietrich, Ph.D. and Peter R. Martin, M.D.

Department of Psychiatry (MSR, PRM) and Department of Biostatistics (MSD), Vanderbilt University School of Medicine, Nashville, TN, USA

Background: Cigarettes and coffee are widely used psychoactive substances among alcoholics. Due to the Keywords: alcohol, Alcoholics Anonymous, cigarette, coffee, devastating public health impact of alcohol use gateway drug disorders, it is important to determine if using cigarettes or coffee may influence alcoholism. Previous studies indicate that cigarette smoking is INTRODUCTION associated with progression of alcohol dependence, Cigarettes and coffee are widely available psychoactive but the effects of coffee drinking have yet to be substances in the USA, typically utilized by individuals investigated. Objectives: To retrospectively determine who also consume beverage alcohol. Due to enormous the temporal sequence of incident cigarette, coffee, healthcare costs and lost productivity associated with and alcohol use and attributed subjective effects in alcohol use disorders and the fact that so many individ- AA participants. Methods: Volunteers at all Nashville uals with these disorders also smoke and drink coffee, = open-AA meetings (n 289 [126 women], completion it is of considerable public health interest to determine = rate 94.1%) were administered a Lifetime Drinking if cigarettes or coffee may influence progression to alco- History modified to also include lifetime cigarette and holism (1,2). coffee consumption, as well as coffee consumption One of the major determinants of subsequent problem- and effects questions, the Fagerstrom Test for atic use of alcohol appears to be the age at which an Nicotine Dependence, and the Smoking Effects individual begins to drink regularly (3). The notion that Questionnaire. Results: Average ages (years) at first the use of a given psychoactive substance (a “gateway regular use of alcohol, cigarettes, and coffee were 15.4 drug”) antedates and predisposes to the use of additional (IQR: 13.0–18.0), 16.7 (IQR: 13.0–18.5), and 18.5 substances and ultimately leads to a diagnosable drug use (IQR: 14.0–23.5), respectively. In a subset who used disorder has heuristic value (4). Cigarettes and coffee are = all three substances (n 236;102 women) alcohol typically used by alcohol-dependent individuals, but the < consumption preceded cigarette smoking (p .001) contribution of these psychoactive substances to problem- < and coffee drinking (p .001), and cigarette smoking atic alcohol drinking is not fully understood despite the < preceded coffee drinking (p .001); these fact that both cigarettes and coffee have been considered relationships did not differ by gender. Conclusions: gateway drugs in their own right. Recovering alcoholics started regular alcohol Findings from epidemiologic studies suggest that consumption prior to cigarette smoking and coffee cigarette smoking frequently precedes alcohol consump- drinking. Scientific Significance: In AA participants, tion in adolescence (4–7) and also enhances subsequent coffee does not precede initiation of regular smoking development of alcohol use disorders (7). Approximately or alcohol drinking as might be anticipated for a 10% of 12-year olds smoked cigarettes within the past gateway drug. 12 months, as did 50% of men and 45% of women

Address correspondence to: Peter R. Martin, Department of Psychiatry, Vanderbilt University School of Medicine, Suite 3068, Vanderbilt Psychiatric Hospital, 1601 23rd Avenue South, Nashville, TN 37232-8650 USA. Tel: +1-(615)-322-3527. Fax: +1-(615)-322-0175. E-mail: [email protected]

27 28 M. S. REICH ET AL. in their early 20s (8). The rapid acquisition of prob- during the summer of 2007 based on review of the AA lematic cigarette smoking is supported by the observa- Central Office website (http://www.aanashville.org/). tion that in 12- to 13-year olds, symptoms of nicotine Meetings were identified by group name, location, time dependence arise less than 2 months after beginning of meeting, and weekday versus weekend. Meetings were monthly cigarette smoking (9). Cigarette and alcohol attended once and in a random order. The chairperson of use are closely linked in the general population and in 75 of the 94 meetings allowed participant recruitment, the those with a drug use disorder (10,11). Between 80 and chairperson of 14 of these meetings chose not to allow 95% of actively drinking alcoholics concurrently smoke recruitment for undetermined reasons, and 5 meetings did cigarettes (11). Furthermore, concurrent out-of-control not actually occur though scheduled due to inadequate use of cigarettes and alcohol may have fundamental neu- attendance. We were unable to identify any apparent robiological underpinnings as suggested by preclinical differences among meetings whose chairperson did and animal studies (12). those that did not allow recruitment of participants. A The potential influence of coffee drinking in devel- total of 307 participants above 18 years, self-identified as opment of alcoholism has not been studied even though a recovering alcoholic, volunteered. We recorded subject coffee is the most widely consumed psychoactive bever- participation and allowed each participant to complete the age in the USA (13), with particularly high prevalence questionnaire on the first occasion only (even if the par- among alcoholics (14). This association is especially ticipant attended multiple meetings and volunteered more important to examine as there is an emerging literature than once). Participants completed questionnaires subse- supporting the beneficial effects of coffee consumption quent to each meeting at the site of the meeting with on various aspects of mental and physical health, includ- one of us (MSR) in attendance to ensure relative quiet ing on complications of chronic alcohol dependence, and attention. A $5 grocery store gift card was offered mediated predominantly by non-caffeine components of as compensation for the time required for participating; this beverage (14–16). Furthermore, findings of previ- approximately 45 minutes were required to complete the ous research are mixed regarding the role of coffee in questionnaire. alcohol dependence. For example, a study of seventh Demographics: Variables were derived from other graders showed that students who drank more than six studies of national drinking patterns (20) and modified for cups of coffee per month were 1.5–2.5 times more likely this study (14). Demographic characteristics included sex, to drink alcohol compared to those who drank six or age, marital status, ethnicity, income, employment sta- fewer cups of coffee per month (17); those who drank tus, religion, endorsed importance of religion, education, more than six cups per month were also more likely to urban/rural status, and state of residence. initiate alcohol consumption within the following year Lifetime Drinking History (LDH): LDH is a validated (4,17). Conversely, another report indicated that coffee and reliable measure of alcohol consumption from onset drinking did not influence alcohol use in teenagers (18). of regular use (21); we modified it to include cigarette Finally, Brazilian youth who drank coffee were found smoking and coffee drinking, explicitly determining at to have a significantly decreased likelihood of develop- what age participants started regularly consuming alco- ing depressive feelings or alcohol cravings compared to hol, tobacco, and coffee. Onset of regular consumption those who did not drink coffee (19). Because little is was defined as when consuming a given substance first known about the relationship between coffee and alco- became habitual or an integral part of daily living (e.g., hol consumption in individuals who subsequently develop the subject indicated daily, weekly, or monthly consump- alcohol use disorders, we retrospectively examined the tion rather than sporadic use). We used fixed 5-year temporal sequence of incident cigarette, coffee, and alco- intervals to measure average daily consumption (22). hol use among a group of Alcoholics Anonymous (AA) Lifetime alcohol use was the sum of values obtained for participants and perceived effects of coffee and cigarette intervals starting at age 11; for those who reported start- use to determine whether coffee might be considered a ing regularly using beer, wine, or liquor before age 11, the gateway drug to alcohol. age uniformly used for analytical purposes was 10 years (8,23). Fagerstrom Test for Nicotine Dependence (FTND): The FTND, developed from the Fagerstrom Tolerance MATERIALS AND METHODS Questionnaire (24), measured nicotine-dependence lev- This research was approved by the Vanderbilt University els in current smokers (25). It is scored on a 10-point Institutional Review Board Behavioral Sciences Commi- scale with dependence differentiated as very low (0–2 ttee and the Middle Tennessee Alcoholics Anonymous points), low, medium (5 points), high, and very high (AA) Central Office in Nashville, TN. One of us (MSR) (8–10 points). met with each AA meeting chairperson to request permis- Smoking Effects Questionnaire (SEQ): This 33-item sion to attend the meeting and to recruit study volunteers. questionnaire assessed personal beliefs regarding the Written informed consent was obtained from each subject effects of smoking (26). Negative effects included neg- before participation. ative physical effects, negative psychosocial effects, and Subjects: Ninety-four different English-speaking, open future health concerns; positive effects included reduced (to the public)-AA meetings occurred in Nashville, TN, negative affect, stimulation, positive social effects, and CIGARETTE AND COFFEE USE IN ALCOHOLISM ONSET 29 weight control. Scoring was based on a 4-point Likert Age of beginning regular cigarette use was approx- scale with 0 – false, 1 – true and hardly important at imately the same for men (N = 145) and women all, 2 – true and moderately important, and3–true (N = 109) (men: M = 16.6 years, median = 15.0, and very important (26). Importance values were defined IQR = 12.0–18.5; women: M = 16.8 years, as the average response value to the items within each median = 15.0, IQR = 13.4–18.5; p = .372). Women subscale (26). began drinking coffee approximately 2 years later than Coffee Effects: Reasons for consuming coffee and men (men: N = 149, M = 18.3 years, median = 18.0, the perceived effects of consumption were measured IQR = 13.5–21.8; women: N = 119, M = 20.1 years, with a questionnaire modified to be gender non-specific median = 18.5, IQR = 15.0–23.5; p = .034). (14,27,28). Sequence of incident use of each drug: Ordering the Statistical Analysis: Statistical summaries and anal- age at which each substance was first regularly used is ysis were performed using SPSS Version 15.0 (SPSS, most reliably ascertained by focusing on subjects who Inc., Chicago, IL) and STATA Version 9.2 (StataCorp, used all three substances. As 82% of men (N = 134) College Station, TX). Descriptive summaries consisted of and 81% of women (N = 102) used all three substances frequency distributions for the nominal and ordinal vari- (Table 3), the results for incident use for each substance in ables. Mean, standard deviation (M±SD), and percentile- this subset of participants were not appreciably different based summaries (median, inter-quartile range (IQR)) from the entire sample indicated above. There were statis- were assessed in cases of severely skewed continuous tically significant differences (p < .001) among the ages variable distributions (e.g., age of onset, consumption). of incident use for the three substances. Incident alcohol The Wilson method was used to calculate confidence use (M = 15.0 years, median = 15.0, IQR = 12.0– intervals for proportions (percentages). This method is 17.0) significantly preceded those of regular cigarette strongly dependent neither on the value of the propor- smoking (M = 16.7 years, median = 15.0, IQR = 13.0– tion nor on sample size; it also does not allow lower 18.5) (p < .001) and coffee drinking (M = 19.0 years, limits to be negative (29). Differences in nominal charac- median = 18.5, IQR = 14.0–23.5) (p < .001). Incident teristics between men and women were summarized via cigarette smoking also preceded incident regular coffee cross-tabulation; tests of statistical significance used chi- use (p < .001). These findings are displayed in terms square tests of independence. Due to the skewed nature of sequence of incident regular substance use (initial, of the continuous data distributions, Mann–Whitney tests middle, or last substance) in Figure 1. The pattern of of statistical significance were used to assess gender dif- incident regular use of substances was not statistically ferences, Friedman tests of ranks was used to test for significantly different for men and women (p = .640). differences in age of beginning to regularly use alcohol, Perceived effects of coffee and cigarettes and severity cigarettes, and coffee. Post hoc analyses of a statisti- of nicotine dependence: Effects of cigarette smoking are cally significant finding were conducted using Wilcoxon summarized in Table 4. Gender differences were detected signed-ranks tests with a Bonferroni-corrected alpha level in negative psychosocial effects (p = .020), reduced neg- of .017. Other than noted for post hoc tests, the maximum ative affect (p = .001), and weight loss benefit (p = .001) alpha level for tests of statistical significance was .05. and were rated more important by women than men. FTND showed both men and women were similarly moderately nicotine dependent (men: M = 5.7, median = 6.0; RESULTS women: M = 6.0, median = 6.0) (p > .05). No rea- Demographics: Volunteers comprising 163 males and son for consuming coffee and no effect of consuming 126 females completed the questionnaire (completion coffee yielded statistically significant gender differences rate: 94.1%). The average age of the participants was (p > .05); these are included in Table 5. 45.1 ± 12.1 years (range 20–82). Men and women were 46.1 ± 11.5 (range 21–79) and 43.8 ± 12.8 (range 20– 82) years old, respectively (p = .028). Other demographic DISCUSSION characteristics are presented in Table 1. Ages at incident substance use: Average overall ages The data retrospectively obtained from our sample of (years) at incident regular use of alcohol, cigarettes, and middle-aged AA participants are consistent with prior coffee among all subjects in the study were 15.4 (IQR: prospective and retrospective studies conducted in young 13.0–18.0), 16.7 (IQR: 13.0–18.5), and 18.5 (IQR: 14.0– adults and adolescents (8,23). Also, lifetime alcohol con- 23.5), respectively. sumption was comparable to data obtained from a clinical Quantity and duration of alcohol consumed by the sample of alcohol-dependent Nashville patients (30). As men and women in this study are summarized in Table presented here, cross-cultural research confirms that men 2. Men tended to begin drinking earlier than women drink larger quantities of alcohol and consume with (men: M = 14.8, median = 15.0, IQR = 12.0–17.0; greater frequency (31,32). Incident alcohol consumption women: M = 16.1, median = 15.0, IQR = 13.0–18.0; in men was 1.3 years earlier than in women (14.8 years p = .086). Men reported a greater number of life- old versus 16.1 years old, respectively) which is compat- time drinks (p < .001) and greater intensity of drinking ible with previous findings showing men who start drink- (drinks/year) (p < .001). ing alcohol before age 15 and women who start before 18 30 M. S. REICH ET AL.

TABLE 1. Demographic variables of participants. Male Female Total Characteristic n (%) n (%) n (%) n 163 (56.4) 126 (43.6) 289 Marital status (n = 288, p = .281) Married 42 (25.8) 29 (23.2) 71 (24.7) Separated 10 (6.1) 10 (8.0) 20 (6.9) Divorced 63 (38.7) 42 (33.6) 105 (36.5) Widowed 1 (.6) 5 (4.0) 6 (2.1) Never married 47 (28.8) 39 (31.2) 86 (29.9) Racial/ethnic background (n = 289, p = .910) White 146 (89.6) 111 (88.1) 257 (88.9) African-American 10 (6.1) 10 (7.9) 20 (6.9) Hispanic 2 (1.2) 2 (1.6) 4 (1.4) Other 5 (3.1) 3 (2.4) 8 (2.8) Income relative to US median∗ (n = 287, p = .040) Same or above 67 (41.4) 37 (29.6) 104 (36.2) Below 95 (58.6) 88 (70.4) 183 (63.8) Employment status∗∗ (n = 289, p = .001) Work full-time 99 (60.7) 56 (44.4) 155 (53.6) Work part-time 18 (11.0) 12 (9.5) 30 (10.4) Retired 12 (7.4) 13 (10.3) 25 (8.7) Homemaker 2 (1.2) 15 (11.9) 17 (5.9) Other 32 (19.6) 30 (23.8) 62 (21.5) Religion (n = 288, p = .070) Catholic 27 (16.7) 14 (11.1) 41 (14.2) Jewish 4 (2.5) 1 (.8) 5 (1.7) Liberal Protestant 21 (13.0) 17 (13.5) 38 (13.2) Conservative Protestant 54 (33.3) 63 (50.0) 117 (40.6) Other 42 (25.9) 25 (19.8) 67 (23.3) None 14 (8.6) 6 (4.8) 20 (6.9) Religious importance∗ (n = 287, p = .045) Very 62 (38.5) 64 (50.8) 126 (43.9) Somewhat 56 (34.8) 34 (27.0) 90 (31.4) Not really 27 (16.8) 11 (8.7) 38 (13.2) Not at all 16 (9.9) 17 (13.5) 33 (11.5) Education level (n = 288, p = .469) Less than high school 7 (4.3) 7 (5.6) 14 (4.9) High school graduate 27 (16.7) 17 (13.5) 44 (15.3) Some college 72 (44.4) 46 (36.5) 118 (41.0) College graduate 54 (33.3) 54 (42.9) 108 (37.5) Other 2 (1.2) 2 (1.6) 4 (1.4) City size/type (n = 283, p = .590) Metro with >50,000 140 (87.0) 107 (87.7) 247 (87.3) Metro with <50,000 9 (5.6) 9 (7.4) 18 (6.4) Non-metro 12 (7.5) 6 (4.9) 18 (6.4) State of residence (n = 284, p = .457) Tennessee 158 (98.8) 121 (97.6) 279 (98.2) Non-Tennessee 2 (1.3) 3 (2.4) 5 (1.8) Note: Abbreviations: n = number. ∗Statistically significant gender difference at p ≤ .05. ∗∗Statistically significant gender difference at p ≤ .001. are more likely to develop alcohol abuse and dependence of recovering alcoholics suggests our findings may be (3). The consistency with the literature on initiation of generalizable beyond this Nashville AA sample (14). regular alcohol consumption as well as observed gender It is of interest that in our sample, incident cigarette differences in alcoholism severity in the present sample smoking did not precede that of regular alcohol drinking CIGARETTE AND COFFEE USE IN ALCOHOLISM ONSET 31

TABLE 2. Alcohol use and duration history. Male Female Total Alcohol age onset (years) (p = .086) n 163 126 289 Mean 14.8 16.1 15.4 SD 3.1 5.1 4.2 Min 10.0 10.0 10.0 25th Pctl 12.0 13.0 13.0 Median 15.0 15.0 15.0 75th Pctl 17.0 18.0 18.0 Max 23.5 44.0 44.0 Consumption years (p = .082) n 162 125 287 Mean 24.6 22.3 23.6 SD 10.3 9.4 10.0 Min 5.8 2.5 2.5 25th Pctl 16.4 15.3 15.6 Median 24.5 22.1 23.8 75th Pctl 32.5 29.5 31.0 Max 49.7 46.9 49.7 Lifetime drinks∗ (p < .001) n 143 113 256 Mean 86, 179.4 61, 850.3 75, 440.4 SD 54, 498.2 57, 034.6 56, 825.8 Min 7, 304.3 1, 526.2 1, 526.2 25th Pctl 40, 761.1 28, 837.2 35, 739.7 Median 76, 094.7 46, 741.4 61, 402.7 75th Pctl 119, 169.7 83, 770.3 105, 694.1 Max 292, 001.7 474, 845.8 474, 845.8 Lifetime kilograms∗ (p < .001) n 143 113 256 Mean 1, 172.0 841.2 1, 026.0 SD 741.2 775.7 772.8 Min 99.3 20.8 20.8 25th Pctl 554.4 392.2 486.1 Median 1, 034.9 635.7 835.1 75th Pctl 1, 620.7 1, 139.3 1, 437.4 Max 3, 971.2 6, 457.9 6, 457.9 Drinking intensity1,∗ (p < .001) n 143 113 256 Mean 3, 664.3 3, 128.3 3, 427.7 SD 1, 998.0 3, 897.7 2, 994.4 Min 231.5 279.7 231.5 25th Pctl 2, 167.5 1, 297.1 1, 658.6 Median 3, 723.2 2, 437.8 3, 154.5 75th Pctl 4, 836.3 3, 926.2 4, 584.5 Max 16, 225.4 39, 465.8 39, 465.8 Note: Abbreviations: n = number; SD = standard deviation; Min = minimum; Pctl = percentile; Max = maximum. ∗Statistically significant gender difference at p ≤ .001. 1Number of drinks per year of drinking. as has been widely reported (4–7). Alcohol was the first of beginning regular cigarette smoking to that of alcohol substance regularly consumed, starting 1.7 years before consumption in the general population or the emergence cigarettes, suggesting that cigarette smoking was not an of dependence to alcohol and nicotine. For example, the antecedent or precipitant of drinking alcohol in our sam- rate of progression from incident use of cigarettes to nico- ple. Consistent with this, it was recently reported that tine dependence is much more rapid than that of incident alcohol promoted the urge to smoke in both men and use of alcohol to alcohol dependence. Therefore, earlier women, a nicotine-seeking effect in men, and effects of (apparent) use of alcohol may belie the fact that out- nicotine and social concomitants in women (33). These of-control cigarette smoking would subsequently actu- findings may or may not have bearing on the relationship ally precede problematic alcohol use. However, it might 32 M. S. REICH ET AL.

TABLE 3. Incident substance use in years of age by gender1. Male (n = 134) Female (n = 102) Total (n = 236) Alcohol age onset (p = .255) Mean 14.6 15.5 15.0 SD 3.0 4.0 3.5 Min 10.0 10.0 10.0 25th Pctl 12.0 13.0 12.0 Median 14.0 15.0 15.0 75th Pctl 17.0 18.0 17.0 Max 21.0 28.5 28.5 Cigarette age onset (p = .439) Mean 16.7 16.9 16.8 SD 6.2 5.6 5.9 Min 10.0 10.0 10.0 25th Pctl 12.0 13.5 13.0 Median 15.8 15.0 15.0 75th Pctl 18.5 18.5 18.5 Max 48.5 38.5 48.5 Coffee use onset (p = .101) Mean 18.3 19.9 19.0 SD 6.1 7.0 6.5 Min 10.0 10.0 10.0 25th Pctl 13.5 14.7 14.0 Median 18.0 18.5 18.5 75th Pctl 22.4 23.5 23.5 Max 38.5 48.5 48.5 Note: Total for subjects who used all substances. Abbreviations: n = number; SD = standard deviation; Min = minimum; Pctl = percentile; Max = maximum.

Substance Alcohol 60.0% Cigarettes Coffee

40.0% Percent 58.5% 51.3%

20.0% 41.5% 36.9% 29.7% 29.2% 21.6% 19.1% 12.3%

0.0% First Middle Last Rank of Age Onset

FIGURE 1. Percentage of subjects and the sequence in which alcohol, cigarettes, and coffee use were initiated. Tied ranks (first, second, third) were assigned the lowest rank (e.g., if age of onset was identical for alcohol and coffee, they both were assigned a rank of “1” and cigarettes assigned “3”). Error bars illustrate 95% confidence intervals of the percentages. also be that individuals with a proclivity for alcohol question posed in this study) is likely differently related dependence initially gravitate to alcohol in preference to to problematic use for different substances. cigarettes. Our data do not specifically address this impor- Our observations concerning initiation of regular cof- tant alternative due to the fact that “regular” use (the fee drinking are important because of the prevalence CIGARETTE AND COFFEE USE IN ALCOHOLISM ONSET 33

TABLE 4. Perceived effects of smoking cigarettes. Red Neg Neg Phys Neg PS Health Affect Stim Pos Social Weight Effect (p = .463) (p = .020)∗ (p = .118) (p < .001)∗∗ (p = .081) (p = .054) (p < .001)∗∗ M1 F2 M1 F2 M1 F2 M1 F2 M1 F2 M1 F2 M1 F2 Mean 1.73 1.83 1.20 1.55 1.59 1.84 2.07 2.45 1.11 1.36 1.24 1.51 .87 1.37 SD .81 .75 .85 .92 1.00 1.02 .77 .60 .84 .87 .84 .85 .86 .92 Median 1.80 2.00 1.20 1.60 1.75 2.00 2.20 2.60 1.00 1.40 1.20 1.40 .50 1.50 Min .00 .20 .00 .00 .00 .00 .00 .80 .00 1.00 .00 .00 .00 .00 Max 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 Note: Abbreviations:NegPhys= negative physical; Neg PS = negative psychosocial; Red Neg Affect = reduced negative affect; Stim = stimulatory; Pos Social = positive social; M = males; F = females; n = number; SD = standard deviation; Min = minimum; Max = maximum. ∗Statistically signiﬁcant gender difference at p ≤ .05. ∗∗Statistically signiﬁcant gender differences at p ≤ .001. 1Males: n = 87. 2Females: n = 67.

TABLE 5. Reasons for consuming coffee and the perceived effects coffee drinking was probably not a risk factor for sub- of consumption. sequent alcohol consumption (or cigarette use) in indi- Men (%) Women (%) viduals who develop alcohol dependence. In this sample of AA participants, both alcohol and cigarette use may Reason for consuming (n = 244) Alert (p = .586) 88.3 86.0 in fact have promoted future coffee consumption (rather Wake up (p = .361) 74.5 79.4 than the obverse), contributing to previously demon- Gets you going in the morning (p = .439) 72.3 76.6 strated positive smoking–coffee and alcohol–coffee cor- Energy (p = .661) 72.3 74.8 relations (10,34,35). Enjoy (p = .628) 70.1 72.9 In individuals who develop alcoholism, alcohol may Taste (p = .941) 65.0 65.4 play a more important role in moderating affect than Stimulates (p = .772) 59.9 61.7 either cigarettes or coffee. While both alcohol and Lift (p = .740) 59.1 57.0 cigarettes elicit their positive perceived effects via the = Habit (p .454) 56.9 61.7 mesolimbic reward pathway (36–38), alcohol may be = Need it (p .392) 39.4 44.9 more rewarding and salient than cigarettes in these indi- Well-being (p = .687) 28.5 30.8 viduals. Alcohol (and cigarettes) may be preferred over Calmsnerves(p = .387) 16.1 12.1 Relax (p = .397) 13.9 10.3 coffee because caffeine (the rewarding constituent of Spouse (p = .451) 5.8 3.7 coffee) does not similarly activate mesolimbic reward Slows you down (p = .759) 2.2 2.8 pathways (39–44). Most participants indicated stimula- Go back to sleep (p = .712) 1.5 .9 tory benefit as reason for drinking coffee and as perceived Effect of consuming (n = 243) effects of consuming coffee, but more than half denied Perk up (p = .567) 88.3 85.8 they “needed it” and more than two-thirds denied it gave Active (p = .325) 78.1 72.6 them a sense of “well-being,” arguing against its role as Efficient (p = .753) 70.8 72.6 an affect modifier to the extent seen with alcohol and = Irritable (p .358) 39.4 45.3 cigarettes. Also, coffee contains other constituents than = Relaxed (p .787) 31.4 33.0 caffeine that may oppose stimulant effects of caffeine and Nervous (p = .967) 23.4 23.6 may modify the psychopharmacological actions of alco- Less efficient (p = .422) 5.8 8.5 Drowsy (p = .814) 4.4 3.8 hol (15,16). In fact, we have previously reported data Sluggish (p = .456) 1.5 2.8 from this same sample suggesting coffee drinking may have a protective effect with respect to sobriety of AA = = Note: Abbreviations: n number; % frequency. participants (14). Limitations of this study pertain primarily to obsta- of coffee consumption and the popular belief that cof- cles related to working with AA participants and with fee may serve as a gateway drug among adolescents. cross-sectional research, but none prevented us from com- Previous research provides arguments both supporting the prehensively studying our sample or reaching legitimate role of coffee as a gateway substance (4,17), and oppos- conclusions. For example, we could only recruit sub- ing it (18,19). While coffee is legally available prior to jects from meetings in which the chairperson allowed us cigarettes and alcohol, here we found that in AA par- to do so. We could not ethically characterize the distin- ticipants, incident coffee drinking is 4.0 and 2.3 years guishing characteristics of meetings to which we were after that for incident alcohol and cigarette consump- not granted access, nor could we interview group mem- tion, respectively. Based on these temporal relationships, bers who chose not to volunteer to participate so as to 34 M. S. REICH ET AL. determine how they differed from those who did volun- active alcoholism and the subsequent likelihood of attain- teer. Therefore, despite these clear ascertainment biases, ing abstinence from alcohol during recovery. Clearly, this we did recruit a sufficiently large and representative sam- approach may have significant impact on understanding ple of volunteers to provide for statistical analyses and the progression of and recovery from this devastating generalizability of results. It may be argued that our sam- disorder. ple may preferentially include more active and involved AA members (45). Also, it is clear that the relationships between incident use, regular use, and abuse and depen- ACKNOWLEDGEMENTS dence for each of these substances are quite different in their rates of progression. However, we see no rea- This study was supported by NIDA (R01 DA015713 and son for these selection biases or differences in trajectory T32 DA021123) and NIAAA (R01 AA014969) grants of drug use to affect the recalled sequences of incident to PRM. We would like to recognize Darcy Roberto use of psychoactive substances among volunteers noted Lima, M.D., Ph.D., for prompting our interest in the in this study or any gender differences. A further lim- health benefits of coffee, Tomas de Paulis, Ph.D., and itation is that we relied on subjects’ self-identification Adriana Farah, Ph.D., for providing the basic chemical as an alcoholic, rather than documenting concurrent or and pharmacologic substrates of this research, A.J. Reid previous inpatient or outpatient treatments, a psychiatric Finlayson, M.D., and Edward F. Fischer, Ph.D., for con- evaluation, or formally assigned DSM IV-TR diagnoses. tributing to our insight into Alcoholics Anonymous (AA) Despite the reliability of LDH in measuring reported demographics and culture, and Nashville’s AA members alcohol use (21) and that this approach has been used for supporting this research and contributing to advancing as the “gold-standard” for studies of other substances of our understanding of alcoholism. abuse for many years, the questionnaire as modified for this study was not formally validated. Thus, it might be argued that discrepancies exist between the actual and Declaration of Interest The authors report no conflict of interest. The authors reported ages of incident use of one or another of the sub- alone are responsible for the content and writing of this stances examined. Given a history of alcoholism and that paper. subjects may tend to focus on their alcohol consumption at AA meetings, alcohol initiation may be more mem- orable than the start of regular cigarette or coffee use. Still, it seems unlikely that a consistent error would be REFERENCES introduced compromising our conclusions. Prior research 1. Grant BF, Stinson FS, Dawson DA, Chou P, Dufour MC, demonstrates there is no reason to suspect that memory of Compton W, Pickering RP, Kaplan K. Prevalence and consuming any particular psychoactive substance would co-occurrence of substance use disorders and independent be impaired to a greater or lesser degree than another sub- mood and anxiety disorders. Arch Gen Psychiatry 2004; 61: stance (46–48). Care must also be taken when interpreting 807–816. findings from cross-sectional surveys, particularly inter- 2. Chen CM, Dufour MC, Yi H. Alcohol consumption among pretations having to do with “time.” We are aware that it young adults ages 18–24 in the United States: results from is impossible to tease out cohort effects (e.g., age and/or the 2001–2002 NESARC Survey. Alcohol Res Health 2004; 28(4):269–281. generational effects) from passage of time effects, as we 3. Dawson DA, Goldstein RB, Chou SP, Ruan JW, Grant BF. have attempted here, in cross-sectional designs. Age at first drink and the first incidence of adult-onset This study is the first, to our knowledge, to objec- DSM-IV alcohol use disorders. Alcohol Clin Exp Res 2008; tively (albeit retrospectively) examine the relation of legal 32(12):119–123. psychoactive substance use to incident use of alcohol in 4. Kandel D. Stages and Pathways of Drug Involvement: a sample of AA participants who ultimately developed Examining the Gateway Hypothesis. Cambridge: Cambridge alcohol dependence. Our data challenge the notion that University Press, 2002. coffee use antedates regular alcohol consumption prior to 5. Torabi M, Bailey W, Majd-Jabbari M. Cigarette smoking as a predictor of alcohol and other drug use by children and ado- development of alcoholism in either men or women who lescents: evidence of the “gateway drug effect”. J Sch Health eventually become alcohol dependent. In spite of show- 1993; 63(7):302–306. ing that neither cigarette nor coffee invariably serves a 6. Parra-Medina D, Talavera G, Elder J, Woodruff S. Role of gateway function for alcohol consumption, we have not cigarette smoking as a gateway drug to alcohol use in Hispanic addressed the important role these substances may have in junior high school students. J Natl Cancer Inst Monogr 1995; accelerating or diminishing alcohol drinking once it has 18:83–86. begun. Nevertheless, this study lays the groundwork for 7. Grucza RA, Bierut LJ. Cigarette smoking and the risk for future research to investigate the psychopharmacological alcohol use disorders among adolescent drinkers. Alcohol effects of cigarettes and coffee on development of alcohol Clin Exp Res 2006; 30(12):2046–2054. 8. Anthony JC, Echeagaray-Wagner F. Epidemiologic analy- dependence, active-disease severity, and recovery. It also sis of alcohol and tobacco use. Alcohol Res Health 2000; lends to the study of how the relationships in sequence of 24(4):201–208. beginning regular alcohol, cigarette, and coffee use ulti- 9. DiFranza JR, Savageau JA, Rigotti NA, Fletcher K, Ockene mately may affect the intensity of alcohol drinking during JK, McNeill AD, Coleman M, Wood C. Development of CIGARETTE AND COFFEE USE IN ALCOHOLISM ONSET 35

symptoms of tobacco dependence in youths: 30 month fol- 26. Rohsenow DJ, Abrams DB, Monti PM, Colby SM, Martin R, low up data from the DANDY study. Tob Control 2002; Niaura RS. The Smoking Effects Questionnaire for adult pop- 11(3):228–235. ulations. Development and psychometric properties. Addict 10. Istvan J, Matarazzo JD. Tobacco, alcohol, and caffeine use: Behav 2003; 28(7):1257–1270. a review of their interrelationships. Psychol Bull 1984; 27. Goldstein A, Kaizer S. Psychotropic effects of caffeine in 95(2):301–326. man. III. A questionnaire survey of coffee drinking and its 11. Patten CA, Martin JE, Owen N. Can psychiatric and chemi- effects in a group of housewives. Clin Pharmacol Ther 1969; cal dependency treatment units be smoke free? J Subst Abuse 10(4):477–488. Treat 1996; 13(2):107–118. 28. Goldstein A, Kaizer S, Whitby O. Psychotropic effects of 12. Steensland P, Simms JA, Holgate J, Richards JK, Bartlett caffeine in man. IV. Quantitative and qualitative differences SE. Varenicline, an alpha4beta2 nicotinic acetylcholine recep- associated with habituation to coffee. Clin Pharmacol Ther tor partial agonist, selectively decreases ethanol consump- 1969; 10(4):489–497. tion and seeking. Proc Natl Acad Sci USA 2007; 104(30): 29. Agresti A, Coull BA. Approximate is better than “exact” for 12518–12523. interval estimation of binomial proportions. The American 13. NCA. National Coffee Drinking Trends. New York, NY: Statistician 1998; 52(2):119–126. National Coffee Association of the U.S.A., Inc., 2007. 30. Parks MH, Dawant BM, Riddle WR, Hartmann SL, Dietrich 14. Reich MS, Dietrich MS, Finlayson AJ, Fischer EF, Martin MS, Nickel MK, Price RR, Martin PR. Longitudinal brain PR. Coffee and cigarette consumption and perceived effects in metabolic characterization of chronic alcoholics with pro- recovering alcoholics participating in alcoholics anonymous ton magnetic resonance spectroscopy. Alcohol Clin Exp Res in Nashville, Tennessee, USA. Alcohol Clin Exp Res 2008; 2002; 26(9):1368–1380. 32(10):1799–1806. 31. Wilsnack RW, Vogeltanz ND, Wilsnack SC, Harris TR, 15. de Paulis T, Schmidt DE, Bruchey AK, Kirby MT, Ahlstrom S, Bondy S, Csemy L, Ferrence R, Ferris J, Fleming McDonald MP, Commers P, Lovinger DM, Martin PR. J, Graham K, Greenfield T, Guyon L, Haavio-Mannila E, Dicinnamoylquinides in roasted coffee inhibit the human Kellner F, Knibbe R, Kubicka L, Loukomskaia M, Mustonen adenosine transporter. Eur J Pharmacol 2002; 442(3): H, Nadeau L, Narusk A, Neve R, Rahav G, Spak F, Teichman 215–223. M, Trocki K, Webster I, Weiss S. Gender differences in 16. de Paulis T, Commers P, Farah A, Zhao J, McDonald MP, alcohol consumption and adverse drinking consequences: Galici R, Martin PR. 4-Caffeoyl-1,5-quinide in roasted coffee cross-cultural patterns. Addiction 2000; 95(2):251–265. inhibits [3H]naloxone binding and reverses anti-nociceptive 32. Dawson DA, Archer L. Gender differences in alcohol effects of morphine in mice. Psychopharmacology (Berl) consumption: effects of measurement. Br J Addict 1992; 2004; 176(2):146–153. 87(1):119–123. 17. Collins LM, Graham JW, Rousculp SS, Hansen WB. Heavy 33. King A, Epstein A, Conrad M, McNamara P, Cao D. Sex caffeine use and the beginning of the substance use onset differences in the relationship between alcohol-associated process: an illustration of latent transition analysis. In smoking urge and behavior: a pilot study. Am J Addict 2008; The Science of Prevention: Methodological Advances from 17(5):347–353. Alcohol and Substance Abuse Research. Bryant K, Windle 34. Klesges RC, Ray JW, Klesges LM. Caffeinated coffee M, West S, eds. Washington, DC: American Psychological and tea intake and its relationship to cigarette smoking: Association, 1998; 79–99. an analysis of the Second National Health and Nutrition 18. Bernstein GA, Carroll ME, Thuras PD, Cosgrove KP, Roth Examination Survey (NHANES II). J Subst Abuse 1994; 6(4): ME. Caffeine dependence in teenagers. Drug Alcohol Depend 407–418. 2002; 66(1):1–6. 35. Swanson JA, Lee JW, Hopp JW. Caffeine and nico- 19. Flores GB, Andrade F, Lima DR. Can coffee help fighting the tine: a review of their joint use and possible interactive drug problem? Preliminary results of a Brazilian youth drug effects in tobacco withdrawal. Addict Behav 1994; 19(3): study. Acta Pharmacol Sin 2000; 21(12):1059–1070. 229–256. 20. Midanik LT, Clark WB. The demographic distribution of US 36. Nestler EJ. Is there a common molecular pathway for addic- drinking patterns in 1990: description and trends from 1984. tion? Nat Neurosci 2005; 8(11):1445–1449. Am J Public Health 1994; 84(8):1218–1222. 37. Di Chiara G. Role of dopamine in the behavioural actions of 21. Skinner HA, Sheu WJ. Reliability of alcohol use indices. The nicotine related to addiction. Eur J Pharmacol 2000; 393(1– Lifetime Drinking History and the MAST. J Stud Alcohol 3):295–314. 1982; 43(11):1157–1170. 38. Kalivas PW, Volkow ND. The neural basis of addiction: a 22. Russell M, Marshall JR, Trevisan M, Freudenheim JL, Chan pathology of motivation and choice. Am J Psychiatry 2005; AW, Markovic N, Vana JE, Priore RL. Test-retest reliability of 162(8):1403–1413. the cognitive lifetime drinking history. Am J Epidemiol 1997; 39. Nehlig A, Boyet S. Dose-response study of caffeine effects 146(11):975–981. on cerebral functional activity with a specific focus on depen- 23. Jernigan D. The USA: alcohol and young people today. dence. Brain Res 2000; 858(1):71–77. Addiction 2005; 100(3):271–273. 40. Di Chiara G. Drug addiction as dopamine-dependent asso- 24. Fagerstrom KO. Measuring degree of physical dependence ciative learning disorder. Eur J Pharmacol 1999; 375(1–3): to tobacco smoking with reference to individualization of 13–30. treatment. Addict Behav 1978; 3(3–4):235–241. 41. De Luca MA, Bassareo V, Bauer A, Di Chiara G. Caffeine 25. Heatherton TF, Kozlowski LT, Frecker RC, Fagerstrom KO. and accumbens shell dopamine. J Neurochem 2007; 103(1): The Fagerstrom Test for Nicotine Dependence: a revision of 157–163. the Fagerstrom Tolerance Questionnaire. Br J Addict 1991; 42. Acquas E, Tanda G, Di Chiara G. Differential effects 86(9):1119–1127. of caffeine on dopamine and acetylcholine transmission 36 M. S. REICH ET AL.

in brain areas of drug-naive and caffeine-pretreated rats. 46. Huerta M, Chodick G, Balicer RD, Davidovitch N, Grotto I. Neuropsychopharmacology 2002; 27(2):182–193. Reliability of self-reported smoking history and age at initial 43. Goldman D, Oroszi G, Ducci F. The genetics of addic- tobacco use. Prev Med 2005; 41(2):646–650. tions: uncovering the genes. Nat Rev Genet 2005; 6(7): 47. Parra GR, O’Neill SE, Sher KJ. Reliability of self-reported 521–532. age of substance involvement onset. Psychol Add Beh 2003; 44. Goldstein A, Kalant H. Drug policy: striking the right balance. 17(3):211–218. Science 1990; 249(4976):1513–1521. 48. Johnson TP, Mott JA. The reliability of self-reported age of 45. Bebbington PE. The efﬁcacy of Alcoholics Anonymous: the onset of tobacco, alcohol and illicit drug use. Addiction 2001; elusiveness of hard data. Br J Psychiatry 1976; 128:572–580. 96:1187–1198. Copyright of American Journal of Drug & Alcohol Abuse is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.