Ahluwalia JS, Harris KJ, Catley D, Okuyemi KS, Mayo MS. Sustained-Release Bupropion for Smoking Cessation in African AmericansA Randomized Controlled Trial. JAMA. 2002;288(4):468–474. doi:10.1001/jama.288.4.468
Author Affiliations: Department of Preventive Medicine and Public Health, University of Kansas School of Medicine, and Kansas Cancer Institute (Drs Ahluwalia, Harris, Catley, Okuyemi, and Mayo); and Departments of Internal Medicine (Dr Ahluwalia) and Family Medicine (Drs Ahluwalia and Okuyemi), University of Kansas School of Medicine, Kansas City.
Context African Americans disproportionately experience greater smoking attributable
morbidity and mortality. Few clinical trials for smoking cessation in African
Americans have been conducted, despite a different profile of both smoking
and quitting patterns.
Objective To compare a sustained-release form of bupropion hydrochloride (bupropion
SR) with placebo for smoking cessation among African Americans.
Design, Setting, and Participants Randomized, double-blind, placebo-controlled trial conducted from February
11, 1999, to December 8, 2000, of 600 African American adults treated at a
community-based health care center. Volunteers, who smoked 10 or more cigarettes
per day were recruited by targeted media and health care professionals.
Intervention Participants were randomly assigned to receive 150 mg of bupropion SR
(n = 300) or placebo (n = 300) twice daily for 7 weeks. Brief motivational
counseling was provided in-person at baseline, quit day, weeks 1 and 3, end
of treatment (week 6), and by telephone at day 3 and weeks 5 and 7.
Main Outcome Measures Biochemically confirmed 7-day point prevalence abstinence at weeks 6
and 26 following quit day.
Results Using intention-to-treat procedures, confirmed abstinence rates at the
end of 7 weeks of treatment were 36.0% in the bupropion SR group and 19.0%
in the placebo group (17.0 percentage point difference; 95% confidence interval,
9.7-24.4; P<.001). At 26 weeks the quit rates
were 21.0% in the treatment and 13.7% in the placebo groups (7.3 percentage
point difference; 95% confidence interval, 1.0-13.7; P
= .02). Those taking bupropion SR experienced a greater mean reduction in
depression symptoms at week 6 (2.96 [9.45] vs 1.13 [8.84]) than those taking
placebo, and after controlling for continuous abstinence, those taking bupropion
SR also gained less weight than those taking placebo.
Conclusions Bupropion SR was effective for smoking cessation among African Americans
and may be useful in reducing the health disparities associated with smoking.
Despite significant reduction in smoking prevalence in the United States,
certain groups retain a higher smoking rate than the general population. African
Americans living in the inner city, for example, have a smoking rate as high
as 45% while the general population's smoking rate is 25%.1,2
The higher smoking rate among this population is not for lack of motivation
to quit. In fact, compared with whites, African Americans are more likely
to attempt to quit smoking more times in any given year.3,4
However, the success rate is 34% lower for blacks than it is for whites.3,4 Studies show that the higher number
of quit attempts would likely result in sustained abstinence if improved smoking
cessation modalities for African Americans are developed and disseminated.5,6
Two randomized clinical trials found a sustained-release form of bupropion
hydrochloride (bupropion SR) to be effective for smoking cessation. In one
study,7 7-day abstinence rates at 6 months
after initiation of the 7-week treatment were 27% for those taking 300 mg
of bupropion SR compared with 16% for those taking placebo. In the other study,8 which assessed bupropion SR in combination with the
nicotine patch, 7-day abstinence rates at 6 months after initiation of 7-week
treatment were 35% for bupropion SR alone, 39% for bupropion SR plus a nicotine
patch, and 19% for the placebo group. Although these studies have shown the
efficacy of bupropion SR to help smokers quit, the subjects were predominantly
white (>93%) and of a middle or higher socioeconomic status (80% with an educational
level of high school or above). The results have limited generalizability
for African Americans, since their smoking patterns are quite different from
those of whites.6 African Americans smoke fewer
cigarettes per day (15 vs 25 for whites), prefer mentholated and higher tar
or nicotine cigarettes,9 and are more likely
to smoke within 10 minutes of awakening.6 Researchers
have also found that African Americans metabolize nicotine slower and have
higher serum cotinine levels per cigarette smoked than do whites.10,11
If the goals of Healthy People 201012
set by the federal government are to be reached, significant smoking cessation
must occur in special populations such as African Americans. Since prior studies
of bupropion SR included few African Americans and few smokers from lower
socioeconomic groups, we conducted a trial to assess the efficacy of bupropion
SR for smoking cessation among inner-city African Americans.
This study was a double-blind, placebo-controlled, randomized trial
of 600 African American smokers. Participants were seen at a community health
care center during a 16-month period and were randomly assigned to receive
a 7-week supply of bupropion SR or placebo. Both groups received 8 counseling
sessions and a previously developed smoking cessation guide.13
Focus groups, a community advisory board, and a small pilot test assisted
in refining the study to maximize cultural sensitivity and enhance recruitment
Participants provided written informed consent during the first visit.
The trial procedures were approved and monitored by the University of Kansas
Medical Center's human subjects committee.
Eligible persons described themselves as being either "African American
or black," were at least 18 years of age, smoked at least 10 cigarettes per
day, were interested in quitting in the next 30 days, spoke English, and had
a permanent home address with a working telephone. Only 1 smoker per household
was allowed to enroll. Participants were excluded if they had a contraindication
for bupropion SR (predisposition to seizures, excessive alcohol use, bulimia
or anorexia nervosa, current use of bupropion, were pregnancy, current use
of psychoactive medication, use of other forms of tobacco or nicotine replacement
in the past 30 days), were in drug treatment during the past 6 months, or
were being treated for depression.
Smokers were recruited using clinic, media, and community outreach,
including speaking at churches and posting signs in local minority-owned businesses.15 Participants were randomized between February 11,
1999, and April 26, 2000, with the last 6-month follow-up session completed
in December 8, 2000. Of the 1498 people who were screened, 981 met screening
criteria and were invited to participate (Figure 1). Sequential enrollment continued until 600 participants
were randomized.15 The randomization codes
were generated in blocks of 50 and sent to the pharmaceutical company, which
packaged the treatment and then shipped the blinded drug to the investigator.
Blinding was successful. At the end of treatment, 58% (150/259) of participants
correctly guessed that they received bupropion SR, and 41% (104/253) correctly
guessed they received placebo. Sixty-one percent (366/600) of participants
attended all medication dispensing visits, a proxy measure for medication
At the baseline visit, participants were randomly assigned to receive
either 150 mg of bupropion SR (Zyban, GlaxoSmithKline, Research Triangle Park,
NC) once a day or placebo once a day for 3 days, after which the dosages were
increased to twice a day for a total of 7 weeks. Study staff instructed all
participants to continue taking pills for the full 7 weeks regardless of their
smoking status. Tablets were packaged in bubble-sheets containing a 1-week
supply. The 45-minute counseling session at the baseline visit included exploring
ambivalence about quitting, preparing to quit, problem solving difficult situations,
and medication adherence. All counselors were African American, had master's
degrees, received training in motivational interviewing,16
and followed semistructured counseling scripts.17
In general, participants received counseling from the same person for all
8 sessions. At the baseline visit, participants also received the smoking
cessation guide Pathways to Freedom13
and completed a battery of assessments.
Participants set a target quit date a week after the start of medication,
and returned on that day (quit day) for a second in-person visit (week 0).
They returned the following week (week 1) and at weeks 3 and 6. These visits
consisted of brief counseling sessions, completion of short batteries of assessments,
and distribution of medications (weeks 1 and 3). Counselors provided brief
supportive telephone calls 3 days after the target quit day and at week 5.
A brief relapse prevention telephone call occurred 7 weeks after the
target quit day, and follow-up assessments occurred at months 3 and 5 (telephone),
and at month 6 (in-person).
Postcard appointment reminders were sent before every visit and participants
who missed appointments were called up to 3 times to reschedule their visit.
For the 6- and 26-week visits staff attempted to call participants as many
as 6 times after having mailed appointment postcards. Participants received
token gifts at every visit (eg, magnet, pen, T-shirt, tote bag, water bottle)
and were reimbursed a total of $100 during the 6-month period: $20 at week
1, $40 at week 6, and $40 at week 26.
The baseline assessment included measures of demographic and health
information, smoking behaviors, and psychosocial variables (Table 1). Nicotine dependence was assessed with the 6-item Fagerström
Test for Nicotine Dependence (FTND) scale.18
Seven withdrawal symptoms from the Diagnostic and Statistical
Manual of Mental Disorders, Fourth Edition19
and cravings were assessed using a modification of the scale developed by
Hughes and Hatsukami.20 Each of 8 symptoms
was rated from 0 (none) to 4 (severe). Psychosocial measures included the
Center for Epidemiologic Studies Depression Scale21
scores, which range from 0 to 60. Scores of 16 or higher indicate the likelihood
of clinical depression.
The primary outcome variable for the study was 7-day point prevalence
smoking cessation at week 26, defined as having smoked no cigarettes—not
even a puff—for the previous 7 days. Secondary outcomes include 7-day
point prevalence smoking cessation at week 6, continuous abstinence at weeks
6 and 26, and change in the number of cigarettes smoked at weeks 6 and 26.
Continuous abstinence was defined as no cigarettes—not even a puff—since
quit day. We allowed participants to miss no more than 1 in-person visit prior
to each assessment. Self-reported abstinence was confirmed with expired carbon
monoxide assessment (≤10 ppm) and discrepancies were resolved by obtaining
saliva for cotinine analysis (≤20 ng/mL). Withdrawal symptoms, nicotine
dependence, and depressive symptoms were also reassessed at weeks 6 and 26.
Surveys were double-data entered and exported into SAS version 8.22 Once the 2 versions of the database were identical,
range and logic checks were performed using SAS statistical software.
The sample size was determined a priori assuming a 2-tailed χ2 test with a type I error of .05, a power of 80%, and a 6-month biochemically
verified (carbon monoxide) quit rate of 14% in the placebo and 24% in the
treatment groups,23 assuming those lost to
follow-up would be imputed as smokers. Under these assumptions 300 smokers
were to be randomly assigned to each group. The 6-month period was chosen
because evidence suggests that cessation rates at 6 and 12 months do not differ
and in a lower socioeconomic status population, it may have been difficult
to follow up all participants for a year.
Categorical baseline variables were summarized by frequencies and percentages,
and quantitative variables were summarized by mean (SD) for each treatment
regimen. Baseline categorical variables were compared using the χ2 test, and quantitative baseline variables were compared using the
2-sample t test.
All statistical analyses were performed on an intention-to-treat basis.
Seven-day point prevalence and continuous abstinence at weeks 1, 3, 6, and
26 were summarized using frequencies and percentages. The χ2
test was used to compare treatments on our primary dichotomous outcome; 7-day
point prevalence at 6 months; and our secondary dichotomous outcomes; 7-day
point prevalence at week 1, week 3, week 6; and continuous abstinence at all
4 times, between treatments. For comparisons on 7-day point prevalence cessation
at weeks 26, 6, 3, and 1, we considered those subjects who failed to return
within their scheduled visit window as smokers. Scheduled visit windows were
3 days before to 7 days after for week 1, 7 days before and after for week
3, 14 days before to 30 days after for week 6, and 30 days before and after
for week 26. Secondarily, for 7-day point prevalence cessation, we also compared
only those who returned regardless of their returning outside of their follow-up
visit interval, and we treated those lost to follow-up as missing. Treating
those lost to follow-up or those being outside the visit window as smokers
provides a conservative estimate but may underestimate the variance, whereas,
analysis on only those who return may overestimate the cessation rate.
We also used generalized estimating equations to conduct longitudinal
analyses comparing verified 7-day point prevalence at weeks 1, 3, 6, and 26
between the 2 groups. This procedure allows inclusion of subjects who returned
outside the appointment window and is appropriate assuming data are missing
completely at random.33,34
Changes in the number of cigarettes smoked were summarized and compared
across groups for all 600 participants imputing no change for those who missed
follow-up or who did not return within scheduled visit windows. Mean (SD)
summarized the change in the number of cigarettes smoked at weeks 6 and 26.
Given the large sample size and the lack of skewness, the 2-sample t test was used to compare the change in the number of cigarettes smoked
at weeks 6 and 26 between the 2 treatment groups.
Frequencies and percentages were used to summarize adverse events by
type over the entire treatment period. Due to the small number of adverse
events, the Fisher exact test was used, for each type of adverse event, to
compare the percentage of subjects who experienced an adverse event between
the 2 groups.
The mean (SD) change in depression scores was summarized at weeks 6
and 26. The 2-sample t test was used to compare the
changes between the 2 groups at weeks 6 and 26. Weight and withdrawal symptom
scores were compared between the 2 treatment groups over time (day 0, week
1, week 3, and week 6) using mixed linear models, assuming an autoregressive
correlation structure, adjusting for baseline (before week 1) levels and whether
a subject had been continuously abstinent from quit day to week 6. Mixed linear
models allow for modeling longitudinal data using fixed and random effects
while taking into account the correlation structure between repeated observations
on individuals over time. This method was chosen over the traditional repeated
measures design to include all subjects in the analysis whether or not they
had returned for all of the visits after quit day.35
There were no significant differences in the baseline characteristics
of participants between treatment groups (Table 1). The average age of the participants was 44 years, and
70% were women. On average, they smoked 17 cigarettes a day, with about half
having previously tried nicotine replacement therapy. Three difference tests
established that the study had high internal and external validity. First,
randomized participants did not differ in sex and number of years they smoked,
but they were older than those who were eligible but did not return for randomization
(44.2 vs 38.8 years). Second, neither treatment allocation nor baseline number
of cigarettes smoked differed between those lost to follow-up and those who
returned at weeks 1, 3, 6, or 26. The only sociodemographic smoking-related
variable listed in Table 1 that
differed was baseline age. Therefore, age at baseline was included as a covariate
in the generalized estimating equation models for 7-day point prevalence abstinence.
Third, for those lost to follow-up at weeks 1, 3, 6, and 26, there were no
differences on these same variables between groups. Of the 600 participants,
470 (78.3%) were followed up at week 6 and 411 (68.5%) at week 26 (Figure 1). For the 189 participants lost
to follow-up at week 26, 107 were seen outside the study window, 81 were not
locatable, and 1 withdrew consent.
The biochemically confirmed point-prevalence abstinence for the treatment
and control groups is shown in Table 2.
The cessation rate for the treatment group was significantly better than the
placebo group at the end of the treatment phase (36.0% vs 19.0%, P<.001) and the improvement was maintained over 26 weeks (21.0%
vs 13.7%, P = .02). These results remained statistically
significant when we analyzed only those subjects who returned, treating those
lost to follow-up as missing. Since our data were determined to be missing
completely at random, we compared the treatment effect longitudinally on 7-day
point prevalence using generalized estimating equations, with and without
controlling for baseline age. When controlling for baseline age, we found
significantly higher rates of smoking cessation in the bupropion SR group
than in the placebo group (OR, 1.19; 95% CI, 1.12-1.26; P<.001). In addition, after assessing the effect of sex on smoking
cessation, we found no sex by treatment interactions nor main effect due to
sex when controlling for treatment.
Rates of continuous abstinence (Figure
2) were significantly higher for the bupropion SR group at weeks
1, 3, and 6 (P<.001 for all comparisons) and at
week 26 (P = .01). Among all participants, the week-6
mean (SD) reduction in number of cigarettes smoked among those taking bupropion
SR was 11.5 (8.4) vs 10.1 (9.1) among those taking placebo (P = .06). At week 26, the reduction among those taking bupropion SR
was 8.3 (10.3) vs 7.9 (8.8) among those taking placebo (P = .61).
We assessed the effects of bupropion SR on withdrawal scores and weight
over 6 weeks using mixed linear models adjusting for baseline values. For
withdrawal scores over time (Figure 3),
there was a significant effect for continuous abstinence (babs = −2.12; P<.001)
but no effect for treatment (bbup = −0.27; P = .40) and no significant interaction effect (b represents the estimated slope parameter). Those who were continuously
abstinent had lower average withdrawal scores compared with those who relapsed.
For weight over time, there were significant differences for continuous
abstinence (babs = 1.25; P<.001) and treatment (bbup
= − 0.62; P = .03) but no significant interaction
effect. Those who were continuously abstinent at week 6 had a significantly
higher average weight over time, and those taking bupropion SR had significantly
lower average weight over time. The differences in weight gained between the
2 groups were not maintained at the week-26 follow-up session.
However, there were differences among the 2 groups in the effect of
bupropion SR on depressive symptoms. Participants who received bupropion SR
experienced a significantly greater mean (SD) reduction in their depressive
symptoms (2.96 [9.45] vs 1.13 [8.84], P = .03) at
6 weeks compared with baseline depressive symptoms. At the week-26 follow-up
session, which is 4½ months after discontinuance of study medications,
this reduction in depressive symptoms was no longer significant (1.56 [9.04]
vs 1.12 [8.74], P = .59).
Table 3 shows the percentages
of participants who reported adverse events at any of the study contact points
through week 6 of drug treatment. Problems sleeping and dry mouth were the
most commonly reported adverse events. There were no reports of rhinitis,
abnormal or bizarre dreams, or seizures. The bupropion SR group was significantly
more likely to report problems sleeping (P = .02).
Twelve participants reported being hospitalized during the treatment phase,
11 of which were known to be unrelated to the medication, and 1 was suspected
to be unrelated to the medication.
To our knowledge, this study is the first to establish the efficacy
of bupropion SR for smoking cessation among African Americans. The 21% quit
rate at 26 weeks for the bupropion SR group was slightly lower than the rates
found in 2 previous studies, which reported 27%7
and 35%8 quit rates among mostly white, middle-class
participants. Our difficulty locating participants within an acceptable time
frame, combined with counting those lost to follow-up as smokers, may underestimate
the actual quit rate. Alternatively, participants in our study may have had
higher relapse rates due to high amounts of stress associated with lower-income
populations.36 However, the 13.7% quit rate
in our placebo group was comparable with the 16%7
and 19%8 quit rates found in these 2 studies.
Higher levels of nicotine addiction among African Americans, as suggested
by some epidemiological and laboratory data,10,11
may partially account for these results. Although lower FTND scores suggest
lower nicotine dependence, the FTND score heavily weights quantity of smoking
and therefore is a less sensitive measure of dependence among lighter smokers,
such as those in our study.37 It has also been
suggested that African Americans develop dependence at lower levels of smoking.6 Finally, the vast majority of participants in this
trial smoked menthol cigarettes. There is some evidence that smokers using
menthol cigarettes may be less likely to quit than those not smoking these
types of cigarettes.38
Our study also found that bupropion SR attenuated weight gained by participants.
Concern about weight gain after smoking cessation is a barrier keeping many
smokers from quitting. Data from the Third National Health and Nutrition Examination
Survey (NHANES III) found that more than half of African American women are
overweight compared with a third of the general population.39
Obesity-related diseases such as diabetes, hypertension, and coronary artery
disease are also more prevalent among African Americans.40
A medication that can help African Americans quit smoking and is also associated
with less weight gain could play an important role for the health of African
Participants who were continuously abstinent from smoking experienced
fewer withdrawal symptoms compared with those who relapsed. However, contrary
to findings from 2 earlier studies, Jorenby et al8
and Dale et al,41 but similar to a third study,
Hurt et al,7 use of bupropion SR did not have
a significant effect on withdrawal symptoms experienced by participants in
our study. This does not support the notion that bupropion SR helps smokers
quit only by reducing their craving or withdrawal of their symptoms.42 Also in contrast to previous studies, participants
who received bupropion SR in our study experienced a significantly greater
reduction in their depressive symptoms compared with those who received placebo.
Although this might be explained by a higher level of depressive symptoms
among participants in our study vs what participants in other studies may
have experienced, the use of different depressive symptom measures precluded
a direct comparison. However, participants in our study had, on average, lower
income and educational levels, and they were less likely to be living with
a partner than participants in other studies. This suggests that there may
be related differences in stress and distress levels that could account for
the discrepant results.
There are a number of strengths of this study. To our knowledge, it
is the largest treatment-based trial of smoking cessation in any ethnic minority
conducted to date. We successfully randomized 600 African American smokers
at a community health care center in a 14-month period. This study also shows
that smokers no matter their economic status want to quit smoking and are
successful. We also achieved good internal and external validity.
There are some limitations to our findings. First, our findings cannot
be generalized to all African Americans. Although this is a community-based
study, it remains a single-site study with certain eligibility criteria such
as the requirement to have access to a telephone for study participation.
Second, similar to our other published work with African American smokers,
the majority of participants were women.22
This may result in a lower overall quit rate in our study populations because
African American women appear to have lower quit rates in the short- and long-term.4 Third, smokers in cessation clinical trials are a
self-selected group who are motivated to quit, which limits the generalizability
but does represent the group of smokers for whom pharmacotherapy may be most
appropriate. Fourth, although our behavioral intervention was more than it
might be in the "real world," it was less intensive than the behavioral interventions
discussed in the other trials using bupropion SR.8
Bupropion SR was effective in assisting African American smokers to
quit. Translating this scientific finding to the population of African American
smokers is critical to further decrease the prevalence of smoking and to reach
the goals of Healthy People 2010. In addition, reductions in smoking will
also aid in reducing the excess smoking-attributable morbidity and mortality
that exists between African Americans and other segments of the population.
Expanding access to medication through health insurance programs, especially
government programs, such as Medicare and Medicaid, may be a wise investment
to reduce some of the health disparities that exist in the United States.