This study examined how race and smoking patterns relate to tobacco dependence among daily smokers (DS) and intermittent smokers (ITS). It found that African American DS reported lower dependence than Caucasian DS, while African American ITS reported higher dependence than Caucasian ITS on some measures, even after controlling for cigarettes per day. The results suggest African American smokers may develop dependence differently based on their smoking patterns. More research is needed to understand racial differences in dependence development and cigarette consumption between smoker types.

1. TOBACCO DEPENDENCE: A
RACE BY SMOKER TYPE
INTERACTION
Jessica Cheng, Dept. of
Epidemiology
Saul Shiffman, Dept. of
Psychology

2. BACKGROUND
•Some data suggest that, among daily smokers (DS), African Americans
develop dependence at lower levels of cigarette consumption than
Caucasians.
•Additionally, African American smokers are more likely to be
intermittent smokers (ITS), who smoke some days but not every day,
than are Caucasian smokers. How race and smoker type relate to
variations in dependence is unclear.
•Here we sought to test how race and smoker type relate to variations
in dependence.

3. METHODS
•Participants (N=482) had to have been
•Smoking for at least 3 years
•Smoking at their current rate for at least 3 months
•Aged 21 years or older
•Participants were community volunteers from the
Pittsburgh area recruited via advertisement
•African American smokers were oversampled because of
their higher likelihood of being ITS

4. DEPENDENCE MEASURES
•Wisconsin Inventory of Smoking Dependence Motives
(WISDM)
•Primary Dependence subscale: Automaticity, Loss of Control, Craving, and
Tolerance
• E.g. “Cigarettes control me”
• Range of scores from 1 to 7
•Secondary Dependence subscale: Affiliative Attachment, Cognitive
Enhancement, Cue Exposure/Associative Processes, Social/Environmental
Goads, Taste, Weight Control, and Affective Enhancement
• E.g. “Cigarettes keep me company like a close friend”
• Range of scores from 1 to 7

5. DEPENDENCE MEASURES (CONTINUED)
•Nicotine Dependence Syndrome Scale (NDSS-T)
•E.g. “After not smoking for a while, I need to smoker to relieve feelings of
restlessness and irritability.”
•Expressed in T-scores where the mean is 50 and standard deviation is 10
•Hooked on Nicotine Checklist (HONC)
•E.g. “Have you ever tried to quit but couldn’t”
•Any score above 0 indicates dependence; Range of scores from 0 to 1
•Time to First Cigarette (TTFC)
•Single item: “How soon after waking up do you smoke your first cigarette?”
•Any time less than 60 minutes indicates dependence
•Expressed via a log transformation

6. ANALYSES
•In this cross-sectional study, a series of linear regression models were used
to assess whether the association between smoker type and multiple
measures of the degree of dependence (as specified) differed by race.
•Smoker type, race, and an interaction term between smoker type and race
were included as independent variables, with control for age, education, and
sex.
•In a second series of models, cigarette consumption, expressed as cigarettes
per day (CPD), was added to evaluate whether this helped explain the
interaction.

7. 14.47
5.85
17.04
4.48
0
2
4
6
8
10
12
14
16
18
Daily Intermittent
Cigarettes Per Day (CPD)
African American Caucasian
Plotted raw means and standard errors CPD is shown.

8. 4.28
2.73
4.62
2.06
1.00
2.00
3.00
4.00
5.00
Daily Intermittent
Wisconsin Inventory of Smoking Dependence Motives
(WISDM)
(Primary Subscale)
African American Caucasian
Plotted raw means and standard errors for the Primary Dependence Motives Subscale of the WISDM is shown. Interaction is significant
(p<.0001) before and after (p=0.0016) adding CPD to the model meaning that while African American daily smokers are less dependent
than Caucasian daily smokers, African American intermittent smokers are more dependent than Caucasian intermittent smokers.

9. 3.90
2.80
4.09
2.67
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
Daily Intermittent
Wisconsin Inventory of Smoking Dependence Motives
(WISDM) (Secondary Subscale)
African American Caucasian
Plotted raw means and standard errors for the Secondary Dependence Motives Subscales of the WISDM. Interaction is not significant
(p=0.0732) before adding CPD to the model nor after (p=0.3751) meaning that any differences between African Americans and
Caucasians are not significantly difference according to smoker type.

10. 45.42
32.57
49.34
29.65
25
30
35
40
45
50
55
Daily Intermittent
Nicotine Dependence Syndrome Scale (NDSS-T)
African American Caucasian
Plotted raw means and standard errors for the NDSS_T. Interaction is significant before (p=0.0020) and after (p=0.0223) adding CPD to
the model meaning that African American daily smokers appear less dependent that Caucasian daily smokers but African American
intermittent smokers appear more dependent than Caucasian intermittent smokers.

11. 0.69
0.50
0.76
0.45
0
0.2
0.4
0.6
0.8
1
Daily Intermittent
Hooked on Nicotine Checklist (HONC)
African American Caucasian
Plotted raw means and standard errors for the HONC. Possible scores range from 0 to 1. Interaction is significant (p=0.02) before adding
CPD to the model but not after (p=0.21) adding CPD meaning that the difference between African American and Caucasian smokers
changes when smoker type is considered but when cigarettes per day is added into the regressions, any differences between African
American and Caucasians remains the same despite smoker type.

12. Plotted raw means and standard errors for time to first cigarette expressed log transformed. Smaller values indicate greater dependence.
Interaction is significant before (p=0.0003) and after (0.0069) adding CPD to the model meaning that the difference in dependence
between African American and Caucasian daily smokers is much less than the difference between African American and Caucasian
intermittent smokers.
2.26
3.88
2.32
5.11
0.00
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
5.00
5.50
Daily Intermittent
Time to First Cigarette (log transformed)
African American Caucasian

13. CONCLUSIONS
•There appears to be a significant interaction between race and smoker type
when it comes to various measures of dependence where African American
smokers are more dependent than Caucasian smokers when looking at
intermittent smokers but less dependent when daily smokers.
•However, cigarettes per day also shows such an interaction, and this
accounts for the effects on some measures of dependence.
•Two composite scales assessing dependence experience–WISDM Primary
Dependence, and NDSS-T– and the time to first cigarette item continued to
show an interaction between race and smoker type even after controlling for
cigarettes per day.

14. CONCLUSIONS (CONTINUED)
•Understanding the nature of dependence is critical to the prevention and treatment of
dependence.
•More research is needed to understand why African American ITS consume more cigarettes
per day than their Caucasian counterparts, and why African American ITS seem to be more
dependent even at similar levels of cigarette consumption.
•Until this research is done, short term solutions may be to try to reduce cigarette
consumption by teaching smokers to limit/ration their smoking or creating tougher laws or
workplace rules that reduce the number of opportunities that smokers have to smoke
cigarettes.
•One lesson that public health officials and treatment specialists may learn from this is that
even if their African American patients are smoking fewer cigarettes than their Caucasian
patients, that does not mean they are less dependent or require less attention or help.

15. FUNDING
•This work was supported by grant R01-DA020742 (Shiffman) from the National
Institutes of Health, National Institute on Drug Abuse.

16. LESSONS LEARNED
•Even cleaned datasets need careful manipulation
• E.g. I changed the coding of African American from those indicating African American race plus any
other race to those indicating African American race and no other race
•It’s important to check that all the variables you need are in a given dataset. I didn’t
do this and then had to combine datasets later
•Adding one extra variable to a regression can drastically change interpretation of
results