Tobacco Documents Online

=~M=M~Mmmmmm mmmi mm Im tSR Nnwack Juurnof of I)ehrn•inrrrl Mcrlicint•. Vol. 12. No. 2. 1989 \\'ilcy, 1.. and Camacho. T. (1980). Lifestyle and future health: Evidencc from the Alantcda County Study. Prey. Mtd. 9: 1-21. \Villiams, A., and \Vechsler, H. (1972). Interrelatedness of preventive actions in health and other a areas. Health Srrv. Rep. 87: 969-976. Vilaliano, P.. Maiuro, R., Rusco, J., and Ilcckcr, .1. (19R7). Itaw vct.u% rclarivc.rnrcc in the asxs%mcnt ol coping stratcgics. J. Uelrav. Med. Itl: 1-18. S•7-ilcrg, N.. \\'cisti. D., and Horowitz, M. (1982). Impact of cvcnt scale: A cru„•vcdid:uion .tudy ' and some empirical evidence supporting a conceptual model of saress rrtiprnt.c syndromes. I J. Consult. Clin. Psrchol. 50: 407-414. Measuring Nicotine Dependence: A Review of the Fagerstrom Tolerance Questionnaire Karl-Olov Fagerstrom" and Nina C. Schneider' In the last decade, the importance of nicotine in maintaining smoking and in cessation difficulty has been acknowledged. Consequently, this has led to efforts to measure nicotine dependence. This paper focuses on a widely used, paper-and-pencil test of nicotine dependence -the Fagerstrom Toler- ance Questionnaire (FTQ. The frndings indicate that the FTQ correlates with other proposed measures of nicotine dependence (carbon monoxide, nico- tnre, and cotinine levels). The connection between FTQ scores and withdrawal ' symptoms is weak. In clinic outcome trials, the FTQ predicted success where no pharmacologic treatment was involved, while nicotine replacement ap- peared to mask the relation between FTQ scores and outcome. However, the FTQ may predict outcome with nicotine replacement as a function of dose. In placebo-controlled, nicotine replacement trials, FTQ scores were related to success by treatment. Problems with the FTQ are described with focus on itenr difficulties and analyses of the scale. KI•:Y \1'OItl)S: nicotine dependence or addiction; Fagerstroin Tolerance Questionnaire; Iobac- co dependence; smoking. INTRODUCTION M66EH09 The smoking of tobacco is a widespread and deeply ;ngrained learned bchavior. With the exception of primary needs, it is probably the next com- 'Phariuacia Leo Therapeutics. Box 941, S-251 09 Hclsingborg. Sweden. 'To whom correspondence should be addressed. rDeparlment of 1 sychiatry and Biobehavioral Sciences, UCLA School of Medicine, Los An- gcles, California 90024, and Psychopharmacology Unit, VA Medical Center, Brentwood T350, 69 1 /11 1 5 1D, Los Angeles, California 90073.

mm~m MM = mmmmm = MM M MM am 160 Fagcrstrom and Schncidcr mon denominator among people of the world. As a learned behavior, what are the reinforcers for tobacco smoking? Why do people continue to smoke in light of devastating hazards to their health? Why is it so difficult to stop 'smoking? 1 [:pcspitc decades of research into motives Ibr tobacco smoking, the qucs- ~0bn of why people smoke is still a challenge. The critical role of nicotine de endence has recently been acknowledged (Surgeon General, 1987). H~wever, it is also clear that the interactions among the psychosocial, sen- „ory, and pharmacologic reinforcing mechanisms are complex (Wes(, 1988) and their rclative roles unknown.]The reinforcement potential of nicotine per se is not simple. Nicotine can reinforce directly (Henningficld et al., 1987; Wesnes and \'Varburton, 1984) and/or through relief of withdrawal induced by deprivation (Hughes et al., 1984; Schneider et al., 1984; Stitzer and Gross, 1988; West, 1984). With increasing empirical evidence for the significant role of nicotine in smoking (Henningfield, 1984), the prevailing point of view is that nico- tine dependence is ultimately the key to the maintenance of smoking and the inability to quit. The focus of this paper is on the Fagerstrom Tolerance Questionnaire (FTQ). The FTQ is an eight-item paper-and-pencil test designed to estimate the degree of nicotine dependence in tobacco smoking. The identification and measurement of nicotine dependence should aid in treatment decisions for tobacco dependence. The FTQ is already being used worldwide, with lit- tle regard paid to its internal validity and consistency, Still, because of its •xidespread use, we review FTQ findings in relation to other measures of dc- pendence and treatment outcome. Are All Smokers Dependent upon Nicotinc? The degree to which nicotine maintains control of smoking will be de- pendent partly upon individual needs. A "social" smoker may be less depen- dent upon nicotine than the smoker who seeks drug-induced changes in affect and performance. A small number of smokers are "chippers," using I or 2 cigarettes a day (Shiffman, in press); others smoke 4-5 packs a day and will smoke when awakened from sleep. The vast majority appears to smoke 10-40 cigarettes a day. It has also been reported that 75-90% of smokers want to quit and have been unable to do so. This daily and frequent pattern of coin- pulsive smoking leads to the categorization of smokers as "dependent." ln contrast, of those who drink alcohol, relatively few are considered alcoholics. That maintaining abstinence is difficult is attested to by high relapse rates reported for all types of treatment programs (Schwartz, 1987). HowMe Nicn/inc Depcndcncc and Ibe Fwgcrslrom Queslionn.ire 161 much failure is due to unrecognized nicotine dependence in smokers seeking help? The position taken by the present authors is that all "regular" smokers may be dependent, in varying degrecs, upon nicotine. If so, there should be a way to assess it. The Fagerstrom Tolerance Questionnaire and other poten- tial measules of dependence (e.g., cotinine), should serve both experimental and applied goals, viz., (I) to identify further the role of nicotine in smok- ing and (2) to provide direction for treatment. An Operational Definition of Dependence For purposes of this paper, dependence on tobacco is defined as com- pulsive use. Compulsive use includes a present state of being unable to quit or stay quit or a past state of difficulty in quitting characterized by with- drawal and/or craving. The latter refers to those who successfully abstain but had great difficulty in doing so. We are not assuming that nicotine dependence accounts for the total dependence upon tobacco. The complexity and interactions of nicotine, be- havioral, and sensory factors in smoking have been well represented c,CVt ~Wk ^sricKpt ~rgIp'cbApnt (see Pomerleau and Pomerleau, 1988). The Measurement of Nicotine Dependence Few direct attempts to measure nicotine dependence among tobacco users are reported in the literature. The full importance of nicotine has been acknowledged only within the last 10 years. In the next few sections, these attempts are briefly summarized. The idea is to look at the directness; and feasibility of other measures and then look at the potential of the FT'Q in replacing time-consuming, expensive, or invasive testing. Cigarette Consumption Initial attempts to assess nicotine dependence assumed that dependence would be a function of the degree of intake and intake could be measured by number of cigarettes. It had been hypothesized that heavy smokers should be more dependent on nicotine than light smokers (e.g., Brantmark et al., 1973) and this should be reflected in outcome. Results were mixed, with heavy smokers less likely to quit smoking in some studies (e.g., Hall ec al., 1984) 6 ot b6 ryot in others (e.g., Fagerstrom, 1982a).

mr.. r .rr r .. M.r. .r i.r r. .. M m... mm..r lm , t )l I 162 Ful;erslrum unJ Schncidcr There are a number of problems with counting the number of cigarettes in the determination of heavy vs. light smoking. The varying nicotine yield of different brands is one. More important, puffing variables such as inhal- ing deeper, holding the smoke longer, and smoking further down to the butt alter the "stated yield" of the cigarette. At the very least, it may be necessary to use an index of the number of cigarettes per day times nominal yield. Even with the use of an "index," we arc far from obtaining a true reflection of nicotine intake. To ascertain nicotine extracted by any individual smoker, we would need to monitor blood nicotine levels at varying intervals. At present, the cost-effective, noninvasive approach of indexing intake through self-report seems most appropriate. Correlations with other measures of in- take will help determine whether there is any validity and reliability to the use of an index. Typologies Several typologies have been constructed to describe smokers accord- ing to when and why they smoke and their ability to refrain from smoking. The most commonly used typologies are the Reasons for Smoking Scale (Ikard and Tomkins, 1973; Tomkins, 1966) and Russell's Classification of Smok- ing by Motives (Russell et al., 1974). The latter is a mix of questions from Ikard et al. (1969) and McKennell (1970). In the Reasons for Smoking Scale, the addiction and negative subscales have been thought to reflect nicotine dependence. The subscales were relat- cd to self-reported craving and withdrawal discomfort in several studics (Ikard and Tomkins, 1973; Lowenthal and Davis, 1976; Williams, 1979) but were not found to be associated with withdrawal in one recent investigation (Hughes and Hatsukami, 1986; see also Hughes, 1984). Russell et al. (1974) subjected Classification of Smoking by Motives to factorial analysis and found a clear-cut separation into two clusters. One included items from the addiction, autonomic, and sedation subscales that could be interpreted as pharmacologically related. The scale has not been subjected to much validity testing. Recent work by Tonnesen et al. (I988a) suggests that a modified scale which the investigators refer to as the Horn- Russell Scale can distinguish those who need nicotine replacement with lower- vs. higher-strength nicotine polacrilex gum. Unfortunately, smokers may be ignorant about nicotine-seeking in their behavior. In the study by Schneider et al. (1983), smokers did not know whether or not they smoked for nicotine or whether or not they smoked for stimulation (Schneider, 1986). Many smokers think that smoking is an over- Nlc4rllnc 1lcpendence mid (lie t'ol;enlrom Quealomurlrc 163 learned behavior with no drug dependence involved. Thus, to assess by self- report, it may be necessary to develop materials that are not dependent upon personal interpretation of behavior but reflect the behavior accurately. This is one of the goals of the FTQ in measuring nicotine dependence. Biochenlical Measures- /ntake Variables In recent years, objective measures of nicotine levels and biochemical checks on abstinence have become standard in the area of smoking research. Carbon monoxide in expired air has been used to validate abstinence in smok- ing cessation and may have potential as a tool in predicting dependence (Fagerstrom, 1982b). Nicotine and cotinine levels are now commonly used to quantify levels of nicotine intake. There is face validity in using nicotine and its main metabolite, cotinine, as measures of nicotine dependence. In one recent study, plasma nicotine levels predicted degree of reported withdrawal (West and Russell, 1985). In another study, precessation plasma cotinine levels predicted outcome in a smoking cessation trial (Hall et al., 1984). THE FAGERSTROM TOLERANCE QUESTIONNAIRE Exactly what characterizes a smoker dependent on nicotine? Can a self- report format, which is easy to administer, be developed? These questions led to the development and testing (Fagerstrom, 1978) of a scale for mcasui~- ing nicotine dependence. It was hypothesized that nicotine dependence should be related to the following: (a) How often the drug was used (number of cigarettes/day). (b) The machine-smoked nicotine yield of the brand. (c) The effective utilization of the drug (through inhalation or not). (d) How soon after awakening and at what rate the drug js used wi- (e) thin the first hour or two after awakening (behavioral self- observation). Plasma nicotine levels should be low in the morn- ing, presumably affecting nicotine-seeking. The first cigarette in the morning. The morning cigarette should be reported as important (perceptual self-observation) because of its power to alleviate supposed nicotine-specific withdrawal symptoms. (f) More internal stimulus control relative to external control. From these six major assumptions, eight questions were constructed to form the Fagerstrom Tolerance Questionnaire (FTQ). V8E66E9V09

r m MMMMMMMMi m MM m mm mum a I r 164 Questions 1. How soon after you wake up do you smoke your first cigarette? 2. Do you find it difficult to refrain from smoking in places where it is forbidden, e.g., in church, at the library, in cinema, etc.? 3. Which cigarette would you hate most to give up? 4. How many cigarettes/day do you smoke? 5. Do you smoke more frequently during the first hours after awakening than during the rest of the day? 6. Do you smokC if you are so ill that you are in bed most of the day? 7. What is the nicotine level of your usual brand of cigarette? R. I)o rou inhale? The FTQ has a scoring range of 0-11 points, with a score of 0 assumed indicative of minimum nicotine dependence and a score of I t indicative of maximum nicotine dependence. The mean score is usually within the range of 5-7 points, with a standard deviation of about 2. For an unselected group of smokers, the mean score is about 5 or 6 (Opinion Research Corp, 1988); for smokers asking for treatment, the mean is usually between 6 and 7 (e.g., Pomerleau et al., in press). The questions and their scoring may need to be adapted to unique cir- cumstances. In some countries, very few may smoke more than 15 cigarettes/day. In other countries, machine-smoked nicotine yields may not be available. The FTQ should also be adapted to changes over lime for op- timal predicting power. For example, if brands with higher yields become less available and fewer smoke high yield brands, the cutoff points may be adjusted. The FTQ (cf. typologies) was designed to assess self-observation of overt behaviors rather than more introspective responses. As noted earlier, sub- jective responses to why an individual smoked can be colored by belief sys- tcros. However, there are two questions in the FTQ (2 and 3) which do require some introspection. These are: "Which cigarette would you hate most to give up?" and "Do you find it difficult to refrain from smoking in places where it is forbidden?" Problems connected with these and other items are discussed below under Adjustments. In addition to the six assumptions, another assumption was entertaincd, t•i;.., that more dependent subjects should have more withdrawal symptoms. Fagerstrom rnd Schneider Answers Points Within 30 min I After 30 min 0 Ycs 1 No 0 Thc first one in the I morning Any other 0 15 or less 0 16-25 1 26 or more 2 Yes 1 No 0 Ycs 1 No 0 0.9 mg or Icss 0 1.0-1.2 mg 1 1.3 mg or more 2 Never 0 Sometimes I Always 2 Nicotinc 1)cpcndencc and the t'agcrstrom Qucsllonnnlrc 165 However, it can be difficult for subjects to answer such a question. Lack of attention to previous withdrawal responses and lack of a previous attempt to quit can interfere with responding. REVIEW OF WORK INCORPORATING THE SCALE In the remainder of the paper, we review the relationship of FTQ scores first to other proposed biochemical indicators of dependence and then to titration, withdrawal, and treatment outcome. Outcome includes cessation trials with and without nicotine replacement. The predictive value of individu- al items and the internal consistency of the scale are also discussed. The FTQ and Biochemical Research Direct measures of nicotine levels or nicotine intake are expected to reflect degree of dependence. In recent years, carbon monoxide, plasma nico- tine, and cotinine levels have been considered in the prediction of outcome. It may be premature to look at the interrelationship among variables not yet established as predictors. However, several investigators have attempted to correlate FTQ and biochemical markers in nonoutcome studies (Table I). Carbon Monoxide and the FTQ Carbon monoxide (CO) levels should reflect, indirectly, nicotine seek- ing and dependence. The higher the CO level, the more puffs or deeper the puffing on cigarettes. As can be seen in Table I, Fagerstrom (1982b) reported a strong corre- lation between CO sampled under normal smoking conditions and scores on the FTQ. Subjects had been selected for even distribution along the FTQ. Statistically significant but not clinically impressive correlations were observed between CO taken at the start of a smoking cessation program and FTQ scores by Killen et al. (unpublished report), Harackiewicz et al. (1988), and Tonnesen et a!. (1988b). Nicotine/Cotinine Levels and the FTQ FTQ scores should also correlate with direct attempts to measure nico- tine and cotinine levels. . ~8E66EMZ

.rr r.r mr.r m r+rr rr a. .. i r. r rm+.r r.r 166 hugcrstram suid ticbneiArr Nicotinc I/cpcnJcncc nnd thc Pngcrstrom Qncs(ionnairc 167 Tehtc t. Iliucltctnicnl Mc:aurc. Study Variahlc" Rcsults/p valuc Fagerstrom (I9R2b) CO & FTQ r = .88,p < .005 Killcn el al. (unpub. data) CO & FTQ r = .1R,p < .(11 llarackicwici M (rl. (1988) cn & I"TQ r 11, /r •- I/I Tonnc.cn et al. (198!(h) CO & PTQ r - .22 p<.(12 a . McNahb (1985) Nicotine & FTQ r . .12. n. 1 I Tonnc.cn (unpuh- data) Tonnc:cn cr al. (19RRb) Nicotine & hi/lo I"TQ Nicotine & FTQ < .(xll r = .23.p <.OI I Tonnc%cn (1988) Lomhardo et al. (1988) Nicotine & FTQ Nicotine boost & FTQ r=.40,p<.nll1 r = .2R, p < -(II Faserstrom (unpub. data) Cotinine & FTQ r=.70,p<.05 Pomcrleau et a!. (1983) Hi/lo cotinine & FTQ -," p < .(125 Hall & I:illcr (1985) I:illcn cr al. (unpub. da(a) Cotininc & ITQ Cotininc & 110 r - .53,p< (tt r .31, p < .ns Tonnesen (unpub. data) Cotininc & IJQ r=.17,ns Pomerleau er al. (in press) Cotinine & <=TQ" r = .33,p<.O01 Colinhtc & FTQ" r = .42, P < .(N)5 Tonncscn (unpub. data) Cotininc & PTQ -," p < -(MII I I 'CO = carbon monoxide taken prior to cessation; nicoline% and cotininc% wcrc also taken prior to cessation except for Lombardo et al. (1988) ((nicotinc boost). "Sec text. In three studies, nicotine levels were sampled at the start of a smoking cessation program and correlated with FTQ scores. The results were mixed (see Table I). McNabb (1985) found no correlation between plasma nicotine and FTQ scores. In contrast, Tonnesen et al. (1988b) and Tonnesen (1988) reported significant correlations between plasma nicotine and the FTQ in both studies. In the second study (Tonnescn, unpublished data), the plasma levels of high-dependent smokers were 28.8 ng/ml, while the low-dependent smokers had nicotine levels averaging 21.2 ng/ml (p < .001). Lombardo et al. (1988) calculated a score for "nicotine boost" defined as the difference in plasma nicotine levels before and following the smoking of a "usual" cigarette. A correlation between "boost" and FTQ scores was significant (see Table 1), with more dependent smokers showing a larger in- crease than lower dependent smokers. There have been eight reports in seven studies examining the relation- ship between cotinine levels and FTQ scores (see Table I). Significant find- ings were found in seven of the eight reports; of the significant findings, five were correlations between cotinine and FTQ, and two were mean com- parisons. Fagcrstrom (unpublished data) looked at cotinine levels while subjects were smoking high- vs. low-yield cigarettes in a titration study. As can be seen, higher levels of cotinine significantly correlated with higher dependence. In the study by Pomcrlcau et al. (1983), the comparisons were based on FTQ scores averaged for smokers already divided into low- and high-cotinine quar- tiles. The mean FTQ scores were 5.3 vs. 8.6 for the two groups, respectively (p < .02). I tall and Killcn (1985) Sampled cotinine prior to smoking cessation and reported a significant correlation between cotinine and FTQ scores. In the study by Killen el al. (unpublished report), cotinine also was sam- pled at the start of a cessation program. In that study, cotinine did not corre- late with withdrawal (another proposed indicator of nicotine dependence). I lowever, higher cotinine levels at the start of cessation correlated significant- ly with higher FTQ scores (Table I). In the same report, thiocyanate taken at the beginning of cessation also correlated positively with the scale (r = .33, p < .05). As can be seen in Table 1, Tonnesen (unpublished data) reported a low and nonsignificant correlation between cotinine and the FTQ. However, in a later study, Tonnesen (unpublished data) reported that subjects with scores of' >7 had cotininc values of 446 ng/ml, which differed significantly from subjects with scores of <7, who averaged 332 ng/ml. Pomerleau et al. (in press) looked at two groups. Among 100 smokers, with no intention of stopping smoking, the investigators reported a signifi- cant correlation (r = .33, p < .001) between cotinine and FTQ. In a smaller sample of intended quitters (N = 50), a significant correlation among these measures was also observed (r =.42, p < .005). It remains to be seen what predictive value the bicchemical markers per se will have for outcome. If they should be of value, a questionnaire which correlates well with those measures offers a convenient, noninvasive, and practical alternative for predicting outcome. It should be noted that in relating FTQ scores to biochemical meas- tires, the role of smoking rate, independent of the scale, has not been consi- dered. The value of the FTQ will depend upon how much more of the variance is accounted for by the scale vs. smoking rate per se (see Internal Charac- teristics of the FTQ, below). ~ Biochemical Markers- Some Cautions In addition to inter- and intrameasurement variability, we must keep in mind what any given measure represents. Plasma levels of nicotine may not be equivalent to dependence upon nicotine. Receptor numbers, accessi- bility, and sensitivity may vary among individuals. Some individuals may be satisfied with relatively low blood nicotine levels yet be very dependent on nicotine. Differences in metabolism and excretion can affect individual responses. 98E66E9f0z 1 .

M M M M M M M M M M ~ M M M M M M M M Prgcrslroin nnd ScUncidcr , 1 I I P We are also assessing nicotine in the venous blood, where concentra- tions may differ from those in areas where nicotine exerts its desired effects (e.g., the CNS, autonomic ganglia, etc.). Psychological choices as to use of nicotine (e.g., in response to stress, etc.) may significantly interact with drug-seeking behavior and thus influence nicotine levels and the other biochemical markers. Physiological Measures, the FTQ, and Nicotine nependence In the study by Fagerstrom (1978) heart rate pre-post smoking a cigarette was used to validate FTQ scores. Heart rate and FTQ showed a significant inverse relation (r = - .69, p < .01). Fagcrstrom (1978) also looked at the relationship between body temperature and FTQ with number of cigarettes partialed out (suggested by others as a key item). In that analy- sis, the correlation was sustained (r = -.54, p < .01). For heart rate and FTQ, when cigarettes per day was partialed out, a minor attenuation was observed (r = -.58, p < .01). Lombardo el al. (1988) have questioned whether the FTQ correlates with heart-rate measures and whether it measures nicotine dependence per se. The investigators looked at five physiological variables: heart-rate peak, heart-rate mean, skin temperature, skin conductance, and blood volume pulse. The measurements were taken pre- and postsmoking of a single cigarette in a standardized manner. A measure of heart-rate "boost" was ad- ded. Four of the seven variables showed no correlation with FTQ scores: The remaining three correlated significantly with FTQ: heart-rate mean (r = .32, p < .05), blood volume pulse (r = .38, p < .05), and nicotine "boost" (r = .28, p < .01). Surprisingly, these correlations were opposite to the ex- pected direction. Interpreting the findings, Lombardo et al. (1988) suggest that the FTQ does not predict tolerance and thus not physical dependence. The investiga- tors conclude that the FTQ "may predict behavioral dependence simply be- cause it identifies smokers who strongly believe that nicotine is an essential part of their smoking habit." However, as noted earlier, many smokers do not believe that they smoke for nicotine and it would be hard to bias an- swers on this scale. Even the two introspective items do not tell the smoker that we are looking for nicotine versus "behavioral" dependence. We should probably await studies that resolve differences between the Fagerstrom (1978) and the Lombardo et al. (1988) findings before drawing any conclusions. C'Iearly, thc relation between physiological variables and I:TQ needs fur- thcr testing. Nicolinc Ucpcndcncc nnd 11ic Ful;erslruni Qucsllonnulrc 169 The FTQ and Titration If titration is a function of nicotine dependence, compensatory fluctu- ations in nicotine levels might also be expected to correlate with FTQ scores and have predictive value for outcome. In the study of Fagerstrom and Bates (1981), subjects smoked two high- and two low-nicotine cigarettes. Puff du- ration, interpuff interval, and number of puffs were used to form a compo- site index.-The more dependent the smoker as assessed by the FTQ, the longer the puff duration, the shorter the interpuff interval, and the greater the num- ber of puffs taken (composite r = .52, p < .05). Also, the more dependent the smoker, the greater the compensation on the low-yield cigarette (r = .53, p < .05). In the study by Fagerstrom (1982a), subjects smoking cigarettes lower in nicotine than their own increased the number of cigarettes. This correlated positively with dependence scores (r = .77, p < .005). Correlation of FTQ Scores and Withdrawal Symptoms Withdrawal symptoms are expected to be, in part, a function of nico- tine dependence. Some of the discomforts associated with withdrawal from smoking may be related to loss of ritualistic behaviors and possibly to loss of sensory reinforcers (e.g., see Rose and Hickman, 1987). However, sever- al studies point to nicotine as the key factor in the appearance of withdrawal (Hughes et al., 1984; Schneider et al., 1984; Stitzer and Gross, 1988; West et al., 1984). Presumably, the more dependent the smoker is on nicotine, the greater the likelihood and/or severity of withdrawal during cessation. Several studies have attempted to correlate measures of withdrawal with FTQ scores at different points in cessation. Fagerstrom (1978) looked at body temperature prior to and two days af- ter stopping smoking. Body temperature decreases (measured orally) correlated inversely with higher scores on the scale (r = -.55, p < .01). Fa#erstrom (1980) also looked at six withdrawal symptoms after 2 weeks of abstinence from smoking. These included headache, dizziness, sweating, difficulties in concentration, constipation, and tingling/creeping sensations beneath the skin. The symptoms were rated by degree of severity and summarized into a withdrawal total. Individuals scoring higher in dependence reported sig- nificantly more withdrawal (r = .40, p < .01). Hughes and Hatsukami (1986) administered a nine-item scale incorporating the following psychological and physiological symptoms: craving, irritability, anxiety, concentrating difficulties, restlessness, hunger, impatience, somat- ic complaints, and insomnia. Withdrawal scores were recorded while still L8E66MOZ

.. .. .. m.. .r +.. M ~r .. +.r .. +.. r mar. , I I 170 Fu{crslrom wnd tichncidcr smoking and three times in the first week of abstinence. The overall with- drawal scores did not correlate with the FTQ scores. By item, craving and difficulty concentrating were significantly higher in those scoring higher on the FTQ (p < .05 for both item correlations). In the study by Killen et al. (unpublished report), an attempt was made to correlate several premeasures-carbon monoxide (CO), cotinine, thiocyanate, cigarette consumption, and FTQ scores-with withdrawal. A 24-item withdrawal scale was given daily for 3 weeks. None of the variables showed any significant correlation with withdrawal. The value of the FTQ and withdrawal was r=.22 (ns). However, in this study, it appears that scores for nicotine gum subjects were combined with scores for placebo sub- jects. Assuming nicotine in gum produced relief, this could reduce any poten- tial correlation when analyzing the data. In general, we may need to look at correlations between withdrawal and FTQ scores within a placebo group for correlations unconfounded by treatment with replacement. In an uncontrolled trial, Gunn (1984) incorporated the FTQ into a clinic setting without nicotine replacement and recorded withdrawal on a 20-item scale. The point in time for assessment is not given other than the author noting that responses were obtained at the fourth session of the American Cancer Society's Fresh Start Program. There was no significant correlation (r =.19, ns) between FTQ scores and overall withdrawal. When the sub- jects were divided into high (_7) and low (<7) scorers, those who scored high on the scale also reported more intense withdrawal (p < .01). The withdrawal studies suggest that there may be a relationship between the experience of withdrawal and FTQ scores, despite the combining of groups and the fact that measures of withdrawal were not taken at the same stages' in cessation across studies. Table It. Overview: Clinics Without Replace- mcnt' Study N Results Fagerstrom (1980) 130 p < .05 Mcintyre (1981) 82 p < .05 p < .07 Jerome el af. (1984) 45 p<.05 Gunn (1984) 80 p < .02 ll & Killen (1985) H 20" a 3 months p < .01 6 months ns 10 months ns DeWit & Camic (1986) 31 No statistics Pinto et a!. (1987) 52 p < .02 'Designs and statistics vary among studics. Sce text for details. 16This was a no-nicotine gum group within a larg- = 8 8£ 6 6 0c9r controlled study. Niculinc I)cpcndcnrc wmd Ihc Fukeraruut Qucalomialrc 171 Where nicotine replacement is incorporated, one immediate advantage of a predictive relationship between FTQ and withdrawal scores would be in the determination of dose requirements (see The FTQ, Degree of Nico- tine Replacement, and Outcome). Smokers scoring higher on the FTQ may need higher doses in nicotine replacement than those who score low. FTQ scores, degree of nicotine replacement, and withdrawal should correlate with outcome. However, we do not yet know the relationship between withdraw- al and short- or long-term abstinence. The FTQ and Outcome Ultimately, it would be desirable to have a scale which predicts out- come and/or identifies who would benefit from a given treatment. In the next sections, the FTQ is reviewed in relation to clinic setting trials (with and without nicotine substitution) and then in relation to findings from placebo-controlled, nicotine-replacement trials. Nicotine replacement, as a direct treatment of nicotine dependence, is expected to influence the rela- tionship between FTQ and outcome. Clinic-Setting Trials The FTQ and Outcome Without Nicotine Substitution The FTQ and outcome have been reported for seven diverse clinical trials which did not incorporate nicotine substitution into treatment. These results are summarized in Table II. In all seven studies, high-dependent scorers were less likely to maintain long-term abstinence than low-dependent scorers. In that nicotine replace- ment directly treats nicotine dependence and may affect outcome, ujing non- substitution or placebo findings should allow us to determine how well the FTQ predicts outcome. In an early clinical trial, Fagerstrom (1980) reported that subjects who relapsed at 3 months had higher FTQ scores than those who remained absti- nent, 7.0 vs. 6.2, respectively (p < .05). In the study by Mclntyre (1981), smoking status at I month correlated with the FTQ (r = .24, p < .05), with more dependent FTQ scorers less likely to remain nonsmokers; a nonsignificant trend in that direction remained at 3 months (r = .17, p < .07). Jerome et al. (1984) studied 45 hospitalized Ml patients. They report- ed that low-dependent scorers were more likely to remain abstinent at 6

.. a... r M .. r sr .r .r .r. ~..r .. M .. M .r .. I.. , I I 172 i I Fagerstrom and Schneider months than high-dependent scorers (p < .05), although exact figures were not reported. Gunn (1984) tested 80 individuals participating in the American Cancer Society's Fresh Start Program. Low FTQ scorers were reportedly more like- ly to remain abstinent than high scorers (p < .02), although neither exact figures nor the time of follow-up testing was given. Within the context of a larger study of nicotine gum vs. a no-nicotine gum control, Hall and Killen (1985) reported correlations for FTQ scores and outcome in the no-nicotine group. At 3 months, low FTQ scorers were more likely to remain abstinent than high scorers (point biserial correlation, N = 20; r = .66, p < .02). No significant association was observed at 6 wccks or at 10 months. DeWit and Camic (1986) provided group treatment for 31 smokers and reported FTQ scores of 8.5 and 6.4 for relapsers and abstainers, respective- ly. In this clinic report, statistics were not reported. Finally, in this catcogry, Pinto et al. (1987) divided 52 smokers in group treatment into those scoring <5 points and >6. Sixty-five percent of low- dependent scorers, vs. 34% of high-dependent scorers, stepped smoking (p < .02). While these clinics used small samples and have validation, time, and control problems, they suggest that the FTQ may predict outcome when nico- tine substitution is not present in treatment. If this is found to be reliable in u•ell-designed, controlled investigations, it would demonstrate (I) the predictive value of the scale and (2) who would be at greater risk without nicotine substitution therapy. The FTQ and Outcome with Nicotine Substitution There are nine trials in clinic settings in which nicotine gum, used ad lib, was included in treatment (see Table 111). The nine trials differed iri set- ting, treatment, and nature of statistical analyses. Of these nine trials, the FTQ was significantly related to outcome in five of the trials and the find- ings were mixed. In the study by Nordenskjold (1983), subjects were allowed three strengths of nicotine gum (1, 2, and 4 mg) and comparisons were made at 4 weeks and 12 months. At 4 weeks, more low-dependent subjects were abstinent than high-dependent subjects (p < .02) but this difference disap- peared at 12 months. From data obtained in the Kornitzer et al. (1987) trial, Kornitzer (per- sonal communication) reported the following: FTQ scores were divided into low, medium, and high ranges for analysis of 3-month dat1. FTQ scores for Nicotine Dependence and the Fagerstrom Questionnaire Table II1. Overview: Clinics With Replacement" Study N Results Nordenskjold (1983) 224 4 weeks p < .02 52 weeks ns Mclin (1984) 117 ns Vanbrabant (1984) 55 ns Edman el a!. (unpub. data) 120 p < .05 Killcn e! al. (unpub. data) 44 ns Harackiewicz el al. (1988) 50 p < .05 DeWit & Camic (1986) 36 ns Kornitz{r el al. (1987) 199 p < .05 Tonncscn el al. (1988b) 107 p < .05 'Dcsigns and statistics vary among studies. See text for details. 173 the three groups were: FTQ scores 0-4 = 63% success, FTQ scores 5-7 = 38010 success, and FTQ scores 8-I 1= 3601o. The difference between the 0-4 and the 5-7 groups was significant (p < .05). Similar findings were observed by Tonnesen et al. (1988b). When sub- jects were divided into >9 vs. <_ 9, 66% of the lower group were abstinent at 3 months, compared to 35% of the very high group (p < .05). Results were similar when subjects were divided into <7 vs. _7. In an earlier study, Melin (1984) reported a nonsignificant trend at I year for low-dependent subjects (median split) to remain abstinent compared to the high-dependent half (30 vs. 40%, respectively) but the trend was not significant. In studies by Edman et al. (unpublished report) and Harackiewicz et al. (1988), the opposite was found. Edman et al. (unpublished report) ran a GP treatment study and found that 38% of the high-FTQ patients were abstinent compared to the low-FTQ patients (2201o) at 1 year (p < .05). In the study by Harackiewicz ef al. (1988), 20% of the high scorers wt're absti- nent vs. 10% of the low-dependent group (p < .05) at I 1 months. And in that study, within a self-help, no-nicotine replacement control group, the higher abstinence rates were observed for the low-dependent group (p < .05). In three of the replacement trials (small samples,), no significant find- ings emerged. Killen et al. (unpublished report) found no difference in out- come as a function of FTQ scores within nicotine gum users in a clinical trial. Vanbrabant (1984) averaged scores at 4 months for abstinent vs. relapsed pa- tients in a GP setting. Abstinent patients averaged 5.7 on the FTQ, vs. 6.4 for nonabstinent patients (ns). DeWit and Camic (1986) averaged scores for 68E66E9voz

M MMM M i M~ MM! M Mi M MMMIM 174 FnRerstrom and Schneider Nicotine Dependence and the Fagerstrom Questionnaire 175 abstainers (7.8) and nonabstainers (8.1) and these scores did not differ sig- nificantly. In the clinical trials, small samples, lack of controls, and major differ- ences in testing and setting may be a problem. In addition, when incorporat- ing nicotine replacement, there is a problem of lack of information on nicotine gum use (dose, frequency, length of use) which can affect outcome. Effec- tive titration of nicotine replacement could mask the predictive value of the FTQ in clinic trials. With placebo controls, we can look at the interaction of replacement vs. no replacement with FTQ high and low scores to deter- mine the questionnaire's efficacy. I Table IV. The FTQ and Outcome: Placebo-controlled, Nicotine Substitution Trials' High dependence Low dependence Follow-uP Study N Nic Plac p Nic Plac p (weeks) Fagerstrom (1982a)" 96 71 39 .05 65 55 ns 26 Jarvik & Schneider (1984) 48 41 8 (.09) 0 30 ns 52 1'agcrstrom (1984) 145 26 51 .05 32 13 (.07) 52 Hall (unpub. data) 120 52 20' .02 58 40 ns 26 Christen el a!. (1984) 193 45 9 .001 28 12 .02 6 Hjalmarson (1984) 195 27 12 (.09) 31 20 ns 52 Hughes & Hatsukami (1985) 100 65 40 (.09) 70 50 ns 2 Dlondal(1987) 187 41 4 .001 46 39 ns 52 Areechon (1987) 199 62 37 .05 52 47 ns 26 Lagrue & Hirsch (unpub. data)d 140 36 18 .05 24 24 ns 26 The FTQ and Outcome: Placebo-controlled, Nicotine Replacement Trials There have been 10 placebo-controlled trials incorporating the FTQ as a predictor of outcome. In Table IV, study subjects were divided into high and low nicotine dependence and then into nicotine replacement vs. placebo conditions. The data were analyzed by percent success within each con- dition. Two of the studies used nonnicotine replacement controls (Fagerstrom, 1984; Hall, 1984). Two studies involved very short-term follow-up (Hughes and Hatsukami, 1985; Christen et al., 1984). All verified abstinence with CO. It is clear from Table IV that nicotine substitution therapy is more ef- fective than placebo for the high-dependent subjects as expected. In the high- dependence category, 7 of the 10 studies show a significant difference with replacement vs. placebo; the remaining 3 show a trend in the same direction. For the low-dependent subjects (with the exceptions of Jarvik and Schneider (1984) and Lagrue and Hirsch (unpublished report)j, there is a trend for nicotine replacement to be more effective than placebo. This is signifi- cant only in the 6-week study by Christen et al. (1984), although the Fager- strom (1984) findings border on significant differences. Looking within the category of nicotine gum use, there are few differ- ences between high- and low-dependent subjects using active gum and the trends are in both directions. Only Jarvik and Schneider (1984) reported a significantly higher success rate for those scoring higher on the scale (41 vs. 0%; p < .05). In general, it is possible that effective titration with nicotine gum in the other studies may have masked the observation of differences within the nicotine replacement groups. Within the category of high- and low-dependent-scoring placebo groups, the trend is more consistent. High-dependent subjects are less successful than low-dependent subjects. As noted earlier, placebo or nonreplacement groups may be best for testing predictions with FTQ scores. The use of such groups eliminates the confound of effective or ineffective replacement. "Ihgh dependcnce, >7; low dependence. <7. Nic, nicotine replacemem; Plae, placebo. "In Fagcrstrom (1982a), high dependence = z8 and low dependence = < 7. Value of 7 were excluded. bNo gum control. "In the study by Lagruc and Hirsch (unpublished data), high dependence =? 8 and low -<8. It should be noted that high dependence is defined conservatively in most of these studies. Except for two studitis (Fagerstrom, 1982a; Lagrue and Hirsch, unpublished report), the cutoff point has been the median or the average (6 to 7). The prediction of success, with substitution, may be more pronounced using the extreme end of the distribution, e.g., scores above 8 or 9 points. Along a similar vein, higher dosing (4 mg or more frequent use of 2 mg) may be more effective in high scorers. The FTQ, Degree of Nicotine Replacement, and Outcome Within the context of no replacement, the FTQ may predict the proba- bility of success. In terms of applied use, the FTQ should not be used to determine who can or cannot receive nicotine replacement (even low scorers benefited). It is suggested that the FTQ could figure in deterrnining what dose/schedule is best for a given individual. Replacement of nicotine with chewing gum-compared to cigarette smoking-is approximately one-third with 2-mg gum and two-thirds with 4-mg gum (Fagerstrom, 1988). Thus, one would predict that high FTQ scorers should do better with the higher dose. Conversely, low FTQ scorers may find the higher dose excessive or aversive and have greater success on the lower dose. Only three studies have compared the efficacy of 2- and 4-mg strengths of nicotine gum. In all three, the degree of nicotine dependence was taken into consideration in analyses of results. nflV9GF9V:17,

M ... r +.. .. M .. M M .. ~ ... M ar. M r .. M W 176 Pngcrstrum rnd Sctmcidcr , t ti Kornitzer et al. (1987) reported 22% abstinence at I ycar for all sub- jects on 2-mg vs. 32% for subjects on 4-mg gum. This difference was not significant. Subjects were then grouped by FTQ scores into low (<5) and high (2t_ 6) dependence. A lower cutoff was used in this trial because the item on nicotine yield (scored 0-2) could not be included. For high-dependent sub- jects, 33% of those who used 4 mg were successful, compared to 18% suc- cess in subjects using 2-mg gum (p < .05). The findings were just the opposite for low-dependent scorers, with 62% success observed for 2-mg subjects com- pared to 29% for those using 4-mg gum, although the latter was not sig- nificant. Tonnesen et a!. (1988b) found no differences in success rates at I year between subjects using 4-mg gum (37%) and those using the 2-mg dose (33%). In contrast to Kornitzer et al. (1987), when analyzed by dependence scores, the high scorers did not do significantly better on 4 mg than the low scorers. In another study by Tonnesen et al. (1988c), high scores (>6) were given either 2- or 4-mg gum; 46% of the high-dependent subjects on 4-mg gum were successful, compared to only 18% of high-dependent subjects using 2-mg gum (p < .05). There is a nonoutcome finding related to dose and FTQ scores. Nor- denskjold (1983) allowed smokers to choose between 2- and 4-mg nicotine gum and observed that higher-dependent smokers on the FTQ opted I'or high doses of gum (r = .20, p < .05). Ultimately, the FTQ and/or other measures of dependence (e.g., pre- quit cotinine levels) may help in the determination of appropriate dose levels in nicotine substitution treatment. At this point, it can only be suggested that the 4-mg strength may be more effective for the higher FTQ scorers. Morc research on this very important issue is needed. Internal Characteristics of the FTQ The main problem with the FTQ is that it has been widely used without proper analysis. We need more information on intcritcm correlations and on the contribution of individual items to the total score, and continued testing of the questionnaire's validity and reliability. With inconsistencies in predicting outcome, we need to know which item may be masking a relationship and which arc the most useful. What power does each item add to the predictive value of the scale? Kabat and Wynder (1987) used one of the proposed "critical" questions in the FTQ-time to the first cigarette in the morning-to distinguish ex- smokers from smokers among a hospital patient population. The present smokers were significantly different from ex-smokers in that the "time to Nicutinc Dcpcndcncc nnJ Ihc Fn{ crslrom Qucslionnafrc 177 first cigarette in the morning" was shorter (p <.01). However, we have no more information on this finding in relation to the overall scale. In general, there are very few data from which to address these issues. Fagerstrom (unpublished data) subjected FTQ data to regression analyses on two samples with 75 subjects per sample. He found that all eight ques- tions were significantly correlated with the total score. The weakest correla- tion, for both samples, was found for the question on inhalation (r = .30 and r = .19). Two other questions had a low correlation with the total score: smoking while ill (r = .37 and r = .46) and having difficulties refraining from smoking in places where it is forbidden (r = .40 and r=.28). For the remaining five items, correlations of each with the total scores ranged between r = .40 and r = .60. The low correlation for inhalation may refelect that the vast majority of smokers inhale; thus, variability is low. Being bedridden may be unusual so that most subjects score this "no." Refraining from smoking in places where it is forbidden is a complicated question. Recent laws have affected many smokers. Smokers may "load up" prior to and immediately following time spent in a restricted environment (e.g., movie threater) and consequently report no difficulty. In the Fagerstrom (unpublished data) study, samples were also subject to regression analyses. In one sample, the order of sum(s) of squares reduced by individual items was as follows: the number of cigarettes per day (Item 4), how soon after awakening (Item 1), and the brand smoked (Item 7). In the second sample, the ranking of items was slightly different: how soon af- ter awakening (Item 1), the number of cigarettes per day (Item 4), and smok- ing more in the morning hours (Item 5). Lichtenstein and Mermelstein (1986) have also investigated the inter- nal structure of the FTQ within a smoking cessation study. The investiga- tors had two samples (N = 179, N = 150) and obtained alpha coefficients for internal consistency of .55 and .51, well below generally accepted stan- dards. In a principal-components factor analysis with varimax rotation in that study, only factor I had an eigenvalue of more than one, 1.6. This fac- tor accounted for 20% of the total variance and consisted of two items: how soon after awakening (Item 1), which loaded .54, and number of cigarettes per day (Item 4), which loaded .25. These two items had a univariate inter- correlation of .40. Factor 2 accounted for 6% of the total variance and was defined by only one item: smoking more in the morning (Item 5), which loaded .55. The factor structure of sample 2 was similar. In Lichtenstein and Mermel- stein (1986), the investigators recommended analyzing individual items and controlling for number of cigarettes (Item 4 on the FTQ). Daily cigarette consumption is an important question, but according to the first author's early work partialing out that question attenuated the as- M66E9V99