Nicotine & Tobacco Research, Volume 16, Number 6 (June 2014) 682–688
Original Investigation
Effect of Cessation Interventions on Hookah Smoking: Post-Hoc Analysis of a Cluster-Randomized Controlled Trial Omara Dogar MPH1, Mohammed Jawad MBBS2, Sarwat K. Shah MSc1, James N. Newell PhD3, Mona Kanaan PhD1, Muhammad A. Khan PhD4, Kamran Siddiqi PhD1,5 1Department of Health Sciences, University of York, York, UK; 2Department of Primary Care and Public Health, Imperial College London, London, UK; 3Nuffield Centre for International Health & Development, Leeds, UK; 4Association for Social Development, Islamabad, Pakistan; 5Hull York Medical School, York, UK
Corresponding Author: Omara Dogar, MPH, Department of Health Sciences, University of York, Seebohm Rowntree Building, Heslington, York YO10 5DD, UK. Telephone: 756-8353670; E-mail: [email protected] Received April 26, 2013; accepted November 23, 2013
Abstract Introduction: We explored the differential effect of cessation interventions (behavioral support sessions with [BSS+] and without [BSS] bupropion) between hookah and cigarette smokers. Methods: We reanalyzed the data from a major cluster-randomized controlled trial, ASSIST (Action to Stop Smoking In Suspected Tuberculosis), which consisted of 3 conditions: (a) behavioral support sessions (BSS), (b) behavioral support sessions plus 7 weeks of bupropion therapy (BSS+), and (c) controls receiving usual care. The trial originally recruited 1,955 adult smokers with suspected tuberculosis from 33 health centers in the Jhang and Sargodha districts of Pakistan between 2010 and 2011. The primary endpoint was continuous 6-month smoking abstinence, which was determined by carbon monoxide levels. Subgroup-specific relative risks (RRs) of smoking abstinence were computed and tested for differential intervention effect using log binomial regression (generalized linear model) between 3 subgroups (cigarette-only: 1,255; mixed: 485; and hookah-only: 215). Results: The test result for homogeneity of intervention effects between the smoking forms was statistically significant (p-value for BSS+: .04 and for BSS: .02). Compared to the control, both interventions appeared to be effective among hookah smokers (RR = 2.5; 95% CI = 1.3–4.7 and RR = 2.2; 95% CI = 1.3–3.8, respectively) but less effective among cigarette smokers (RR = 6.6; 95% CI = 4.6–9.6 and RR = 5.8; 95% CI = 4.0–8.5), respectively. Conclusions: The differential intervention effects on hookah and cigarette smokers were seen (a) because the behavioral support intervention was designed primarily for cigarette smokers; (b) because of differences in demographic characteristics, behavioral, and sociocultural determinants; or (c) because of differences in nicotine dependency levels between the 2 groups.
Introduction Hookah tobacco smoking, also known as water pipe, shisha, narghile, or hubble bubble, is a 600-year-old apparatus used to smoke tobacco through water (Chattopadhyay, 2000). This once-culturally confined practice is now gaining popularity, particularly among the youth in selected regions worldwide. For example, prevalence among university students in the United States ranges from 15% to 20% (Akl et al., 2011), and in Pakistan, it is as high as 33% (Jawaid et al., 2008). Described as a second tobacco epidemic following that caused by cigarettes (Maziak, 2011), reasons for its continuing growth are multifactorial but broadly rooted in an erroneous belief that hookah is safer than cigarettes (Akl, Jawad, Lam, Obeid, & Irani, 2013). Despite a paucity of high-quality studies, a recent systematic review identified that hookah smoking is significantly
associated with lung cancer, respiratory illness, periodontal disease, and low birth weight in newborns (Akl et al., 2010); moreover, acute concerns surround the potential for carbon monoxide (CO) poisoning (Fauci, Weiser, Steiner, & Shavit, 2012). Other analyses have identified no difference between cigarette and hookah smokers’ lung function (Boskabady, Farhang, Mahmodinia, Boskabady, & Heydari, 2012; Raad et al., 2011), suggesting that hookah use can lead to chronic pulmonary obstructive disease (Salameh et al., 2012). Brief and opportunistic advice by a health care professional is an effective cessation support for patients who smoke (NICE, 2008). Similarly, intensive behavioral support alone or in combination with pharmacotherapies is also effective in promoting smoking cessation (Lancaster & Stead, 2005). The “active ingredients” of the behavioral support interventions typically involve directly assessing motivation, maximizing
doi:10.1093/ntr/ntt211 ©Advance Access publication December 27, 2013 © The Author 2013. Published by Oxford University Press on behalf of the Society for Research on Nicotine and Tobacco. All rights reserved. For permissions, please e-mail: [email protected].
682
Nicotine & Tobacco Research self-regulatory capacity, promoting adjuvant activities, and supporting other behavior change techniques (BCTs) (Michie, Hyder, Walia, & West, 2011). Pharmacotherapies, such as use of nicotine replacement therapy (Stead, Perera, Bullen, Mant, & Lancaster, 2008), bupropion (Hughes, Stead, & Lancaster, 2007), and varenicline (Cahill, Stead, & Lancaster, 2012), substantially increase the likelihood of successful smoking cessation compared to placebo use. However, all these therapies proven effective for smoking cessation have been only evaluated among cigarette smokers. We did not find published effectiveness studies of any tobacco cessation intervention targeting hookah smokers (Maziak, Ward, & Eissenberg, 2007). In this article, we reanalyze the data from a major smoking cessation cluster-randomized controlled trial, Action to Stop Smoking In Suspected Tuberculosis (ASSIST), conducted in Pakistan (Siddiqi et al., 2013; Siddiqi, Khan, Ahmad, & Shafiq-ur-Rehman, 2010). The trial included hookah as well as cigarette smokers; so, we took the opportunity to conduct a post-hoc analysis on the ASSIST data to explore whether there is a difference in the effect of the cessation interventions between hookah and cigarette smokers.
Methods Thirty-eight primary and secondary health care centers registered as diagnostic centers by the tuberculosis (TB) program in Jhang and Sargodha districts of Pakistan were invited to participate in the ASSIST trial between 2010 and 2011; 33 centers agreed to participate. The study enrolled 1,955 patients aged 18 years or older with suspected pulmonary TB (cough for >3 weeks without any other cause) who were also regular tobacco smokers (>1 cigarette/hookah session a day). Those who consented to participate were randomized to three conditions: patients in one group received two brief behavioral support sessions (BSS group), patients in the second group received two brief BSS plus 7 weeks of bupropion therapy (BSS+ group), and patients in the control group received usual care. Given the lack of any routine advice or educational materials in Pakistan, we provided a self-help leaflet on smoking cessation to all participants (for details, refer to the Supplement2, available at www.annals.org; Siddiqi et al., 2013). Patients requiring hospitalization or urgent medical attention were excluded. The support sessions consisted of two structured consultations based on the World Health Organization’s “5 As Approach” (for details, refer to the Supplement1, available at www.annals.org; Siddiqi et al., 2013). The aim of the first consultation was to assist smokers who were willing to set a quit day (a week after the first contact) by encouraging them to see themselves as nonsmokers, planning for their quit day, and preparing them for the initial stages of the quit attempt, which was planned to be 30 min long. The second consultation coincided with the quit date of the patient and provided an opportunity for follow-up and review progress consisting of 10 min in duration. The activities within each of these consultations were designed using BCTs, considered effective in smoking cessation (Michie et al., 2011). In addition to BSS, participants in the BSS+ group also received sustained-release bupropion, 75 mg/day for the first week and 150 mg/day thereafter. They were asked about any adverse effects and treatment adherence, the latter of which was based on self-reports at follow-up. The intervention was delivered by TB DOTS (directly observed
treatment, short-course) paramedics, who received one fullday training on intervention protocol and delivery tools. These DOTS paramedics are nurses or auxiliary nurses who follow a physician’s clinical directions on TB treatment and record patient progress in TB registers, monitor direct administration of TB medications, and ensure clinical follow-up of patients in the health centers. The primary outcome was continuous smoking abstinence, defined as an expired CO measurement (piCO+Smokerlyzer, Bedfont Scientific) of 9 ppm or less (Russell standard) (West, Hajek, Stead, & Stapleton, 2005) at the 5- and 25-week postintervention follow-up visits. Hookah smokers exhibit many of the same features of nicotine dependency as cigarette smokers (Auf et al., 2012); thus, the Fagerström Test for Nicotine Dependency (FTND; now called Fagerström Test for Cigarette Dependence; Fagerström, 2012) was adapted for use in the ASSIST trial. Both treatment conditions led to a eight- to ninefold increase in continuous smoking abstinence compared to the result with usual care; more details can be found in the published protocol and findings of this trial (Siddiqi et al., 2010, 2013). Majority (1,255) of the participants in the trial smoked cigarettes, 485 smoked both hookah and cigarettes (mixed group), and about 215 smoked hookah exclusively. All hookah smokers used unflavored tobacco in which tobacco leaves are mixed with a by-product of cane sugar purification to prepare a tobacco/molasses mixture containing equal proportions of both by weight. This is much similar to Arabian jurâk, which contains pulpy fruit in addition to tobacco and molasses. Participant characteristics at enrollment (like age, gender, duration of smoking, nicotine dependence, and CO reading) were compared between the three smoking-form subgroups using formal statistical tests (for mean: Tukey’s least significant difference test; for median: Kruskal–Wallis test; and for nominal: chi-square test). Subgroup-specific relative risks (RRs) of smoking abstinence and CIs were computed and tested for differential intervention effect among the three forms of smoking using homogeneity test statistics (likelihood ratio estimates for Type 3 analysis) (Assmann, Pocock, Enos, & Kasten, 2000; Lagakos, 2006; Wang, Lagakos, Ware, Hunter, & Drazen, 2007) in SAS, version 9.3 (SAS Institute) using Proc Genmod (generalized estimating equations models). Estimates adjusted for the covariates—age, gender, and duration of smoking—are also presented. These adjusted estimates (adj. RRs) are used to construct a forest plot, with a log RR scale and marker size corresponding to sample weight in each subgroup, for further visual interpretation. It is very important to realize that the effect size estimates and CIs in these analyses are not presented to show statistical significance; rather, they are interpreted as providing a plausible range of intervention differences, consistent overall with the main trial results (Lagakos, 2006).
Results The primary outcome data were missing for 5.8% of the total sample, that is, 15 hookah smokers, 74 cigarette smokers, and 25 mixed smokers. There was no differential loss to followup or intervention completion among the three subgroups (Figure 1). There were significant differences among the three subgroups at enrollment: hookah-only smokers were older—with
683
Behavioral support for hookah smokers Trial participants in total: 1955
Subgroups of smoking form
Hookah BS+: 27 BS: 118 Control: 70 Total: 215
Cigarette
Mixed
BS+: 465 BS: 371 Control: 419 Total: 1255
BS+: 167 BS: 151 Control: 167 Total: 485
Did not complete intervention BS+: 4 BS: 1
BS+: 28 BS: 6
BS+: 8 BS: 1
Loss to follow-up BS+: 7 BS: 2 Control: 6 Total: 15
BS+: 37 BS: 13 Control: 24 Total: 74
BS+: 9 BS: 5 Control: 11 Total: 25
Analyzed BS+: 20 BS: 116 Control: 64 Total: 200
BS+: 428 BS: 358 Control: 395 Total: 1181
BS+: 158 BS: 146 Control: 156 Total: 460
Figure 1. Flow chart of subgroups of smoking forms enrolled. more females, higher CO readings, lower nicotine dependency score, and longer duration of smoking—than the cigarette-only smokers (Table 1). The test for homogeneity of intervention effects among these three subgroups was statistically significant (p-value for BSS+: .04 and that for BSS: .02), both for the crude and the adjusted analyses (Table 2). Compared to the control, the RR of smoking abstinence for BSS+ was 2.5 (95% CI = 1.3–4.7) and that for BSS was 2.2 (95% CI = 1.3–3.8) among hookah-only smokers; and the RR for BSS+ was 6.6 (95% CI = 4.6–9.6) and that for BSS was 5.8 (95% CI = 4.0–8.5) among cigarette-only smokers(Table 2). After adjustment for age, gender, and duration of smoking, the effect estimates essentially remained the same (Table 2).
Discussion These findings suggest that behavioral support with and without bupropion appears to be effective in achieving 6-month
684
smoking abstinence among hookah-only smokers but to a lesser extent than among cigarette-only smokers. It is possible that the differences observed among the three subgroups might be (a) due to the variation in characteristics of the individuals, the way they smoke (habits regarding puff depth and frequency), and sociocultural determinants or (b) due to differences in nicotine dependency levels. The BCTs used in our intervention were primarily developed to influence behavioral determinants of cigarette smoking. However, unique behaviors and stimuli connected to the nicotine delivery products are likely to differ across the various forms of smoking, particularly hookah smoking. Several potential limitations of this post-hoc analysis should be considered. Tests of interaction typically have low statistical power, and the sample size calculation for this trial did not take into account the power to detect heterogeneity of intervention effects by smoking forms. However, there were sufficiently large numbers within each subgroup to allow adequate analysis. Risk of false-positive results increases with multiple subgroup testing; therefore, the analysis was restricted to the
Nicotine & Tobacco Research Table 1. Participant Characteristics by Subgroups of Smoking Form at Enrollment: ASSIST Pakistan Trial Characteristic Mean age (SD) in years Number of females (%) Median household income per month (IQR) in USD Median duration of smoking (IQR) in years Median smoking start age in years (IQR) Median quantity smoked daily (IQR)a Sample (interventions only) Mean CO reading at recruitment (SD) Mean Fagerström score (SD) Number (%) of patients by severity of nicotine dependency Very high High Medium Low Very low
Cigarette smokers (n = 1,255)
Mixed smokers (n = 485)
Hookah smokers (n = 215)
p value
37.4 (11.5) 21 (1.7) 81.4 (58.1)
45.4 (13.3) 24 (5.0) 93.0 (81.4)
51.5 (13.8) 45 (21.0) 81.4 (69.8)
Original Investigation
Effect of Cessation Interventions on Hookah Smoking: Post-Hoc Analysis of a Cluster-Randomized Controlled Trial Omara Dogar MPH1, Mohammed Jawad MBBS2, Sarwat K. Shah MSc1, James N. Newell PhD3, Mona Kanaan PhD1, Muhammad A. Khan PhD4, Kamran Siddiqi PhD1,5 1Department of Health Sciences, University of York, York, UK; 2Department of Primary Care and Public Health, Imperial College London, London, UK; 3Nuffield Centre for International Health & Development, Leeds, UK; 4Association for Social Development, Islamabad, Pakistan; 5Hull York Medical School, York, UK
Corresponding Author: Omara Dogar, MPH, Department of Health Sciences, University of York, Seebohm Rowntree Building, Heslington, York YO10 5DD, UK. Telephone: 756-8353670; E-mail: [email protected] Received April 26, 2013; accepted November 23, 2013
Abstract Introduction: We explored the differential effect of cessation interventions (behavioral support sessions with [BSS+] and without [BSS] bupropion) between hookah and cigarette smokers. Methods: We reanalyzed the data from a major cluster-randomized controlled trial, ASSIST (Action to Stop Smoking In Suspected Tuberculosis), which consisted of 3 conditions: (a) behavioral support sessions (BSS), (b) behavioral support sessions plus 7 weeks of bupropion therapy (BSS+), and (c) controls receiving usual care. The trial originally recruited 1,955 adult smokers with suspected tuberculosis from 33 health centers in the Jhang and Sargodha districts of Pakistan between 2010 and 2011. The primary endpoint was continuous 6-month smoking abstinence, which was determined by carbon monoxide levels. Subgroup-specific relative risks (RRs) of smoking abstinence were computed and tested for differential intervention effect using log binomial regression (generalized linear model) between 3 subgroups (cigarette-only: 1,255; mixed: 485; and hookah-only: 215). Results: The test result for homogeneity of intervention effects between the smoking forms was statistically significant (p-value for BSS+: .04 and for BSS: .02). Compared to the control, both interventions appeared to be effective among hookah smokers (RR = 2.5; 95% CI = 1.3–4.7 and RR = 2.2; 95% CI = 1.3–3.8, respectively) but less effective among cigarette smokers (RR = 6.6; 95% CI = 4.6–9.6 and RR = 5.8; 95% CI = 4.0–8.5), respectively. Conclusions: The differential intervention effects on hookah and cigarette smokers were seen (a) because the behavioral support intervention was designed primarily for cigarette smokers; (b) because of differences in demographic characteristics, behavioral, and sociocultural determinants; or (c) because of differences in nicotine dependency levels between the 2 groups.
Introduction Hookah tobacco smoking, also known as water pipe, shisha, narghile, or hubble bubble, is a 600-year-old apparatus used to smoke tobacco through water (Chattopadhyay, 2000). This once-culturally confined practice is now gaining popularity, particularly among the youth in selected regions worldwide. For example, prevalence among university students in the United States ranges from 15% to 20% (Akl et al., 2011), and in Pakistan, it is as high as 33% (Jawaid et al., 2008). Described as a second tobacco epidemic following that caused by cigarettes (Maziak, 2011), reasons for its continuing growth are multifactorial but broadly rooted in an erroneous belief that hookah is safer than cigarettes (Akl, Jawad, Lam, Obeid, & Irani, 2013). Despite a paucity of high-quality studies, a recent systematic review identified that hookah smoking is significantly
associated with lung cancer, respiratory illness, periodontal disease, and low birth weight in newborns (Akl et al., 2010); moreover, acute concerns surround the potential for carbon monoxide (CO) poisoning (Fauci, Weiser, Steiner, & Shavit, 2012). Other analyses have identified no difference between cigarette and hookah smokers’ lung function (Boskabady, Farhang, Mahmodinia, Boskabady, & Heydari, 2012; Raad et al., 2011), suggesting that hookah use can lead to chronic pulmonary obstructive disease (Salameh et al., 2012). Brief and opportunistic advice by a health care professional is an effective cessation support for patients who smoke (NICE, 2008). Similarly, intensive behavioral support alone or in combination with pharmacotherapies is also effective in promoting smoking cessation (Lancaster & Stead, 2005). The “active ingredients” of the behavioral support interventions typically involve directly assessing motivation, maximizing
doi:10.1093/ntr/ntt211 ©Advance Access publication December 27, 2013 © The Author 2013. Published by Oxford University Press on behalf of the Society for Research on Nicotine and Tobacco. All rights reserved. For permissions, please e-mail: [email protected].
682
Nicotine & Tobacco Research self-regulatory capacity, promoting adjuvant activities, and supporting other behavior change techniques (BCTs) (Michie, Hyder, Walia, & West, 2011). Pharmacotherapies, such as use of nicotine replacement therapy (Stead, Perera, Bullen, Mant, & Lancaster, 2008), bupropion (Hughes, Stead, & Lancaster, 2007), and varenicline (Cahill, Stead, & Lancaster, 2012), substantially increase the likelihood of successful smoking cessation compared to placebo use. However, all these therapies proven effective for smoking cessation have been only evaluated among cigarette smokers. We did not find published effectiveness studies of any tobacco cessation intervention targeting hookah smokers (Maziak, Ward, & Eissenberg, 2007). In this article, we reanalyze the data from a major smoking cessation cluster-randomized controlled trial, Action to Stop Smoking In Suspected Tuberculosis (ASSIST), conducted in Pakistan (Siddiqi et al., 2013; Siddiqi, Khan, Ahmad, & Shafiq-ur-Rehman, 2010). The trial included hookah as well as cigarette smokers; so, we took the opportunity to conduct a post-hoc analysis on the ASSIST data to explore whether there is a difference in the effect of the cessation interventions between hookah and cigarette smokers.
Methods Thirty-eight primary and secondary health care centers registered as diagnostic centers by the tuberculosis (TB) program in Jhang and Sargodha districts of Pakistan were invited to participate in the ASSIST trial between 2010 and 2011; 33 centers agreed to participate. The study enrolled 1,955 patients aged 18 years or older with suspected pulmonary TB (cough for >3 weeks without any other cause) who were also regular tobacco smokers (>1 cigarette/hookah session a day). Those who consented to participate were randomized to three conditions: patients in one group received two brief behavioral support sessions (BSS group), patients in the second group received two brief BSS plus 7 weeks of bupropion therapy (BSS+ group), and patients in the control group received usual care. Given the lack of any routine advice or educational materials in Pakistan, we provided a self-help leaflet on smoking cessation to all participants (for details, refer to the Supplement2, available at www.annals.org; Siddiqi et al., 2013). Patients requiring hospitalization or urgent medical attention were excluded. The support sessions consisted of two structured consultations based on the World Health Organization’s “5 As Approach” (for details, refer to the Supplement1, available at www.annals.org; Siddiqi et al., 2013). The aim of the first consultation was to assist smokers who were willing to set a quit day (a week after the first contact) by encouraging them to see themselves as nonsmokers, planning for their quit day, and preparing them for the initial stages of the quit attempt, which was planned to be 30 min long. The second consultation coincided with the quit date of the patient and provided an opportunity for follow-up and review progress consisting of 10 min in duration. The activities within each of these consultations were designed using BCTs, considered effective in smoking cessation (Michie et al., 2011). In addition to BSS, participants in the BSS+ group also received sustained-release bupropion, 75 mg/day for the first week and 150 mg/day thereafter. They were asked about any adverse effects and treatment adherence, the latter of which was based on self-reports at follow-up. The intervention was delivered by TB DOTS (directly observed
treatment, short-course) paramedics, who received one fullday training on intervention protocol and delivery tools. These DOTS paramedics are nurses or auxiliary nurses who follow a physician’s clinical directions on TB treatment and record patient progress in TB registers, monitor direct administration of TB medications, and ensure clinical follow-up of patients in the health centers. The primary outcome was continuous smoking abstinence, defined as an expired CO measurement (piCO+Smokerlyzer, Bedfont Scientific) of 9 ppm or less (Russell standard) (West, Hajek, Stead, & Stapleton, 2005) at the 5- and 25-week postintervention follow-up visits. Hookah smokers exhibit many of the same features of nicotine dependency as cigarette smokers (Auf et al., 2012); thus, the Fagerström Test for Nicotine Dependency (FTND; now called Fagerström Test for Cigarette Dependence; Fagerström, 2012) was adapted for use in the ASSIST trial. Both treatment conditions led to a eight- to ninefold increase in continuous smoking abstinence compared to the result with usual care; more details can be found in the published protocol and findings of this trial (Siddiqi et al., 2010, 2013). Majority (1,255) of the participants in the trial smoked cigarettes, 485 smoked both hookah and cigarettes (mixed group), and about 215 smoked hookah exclusively. All hookah smokers used unflavored tobacco in which tobacco leaves are mixed with a by-product of cane sugar purification to prepare a tobacco/molasses mixture containing equal proportions of both by weight. This is much similar to Arabian jurâk, which contains pulpy fruit in addition to tobacco and molasses. Participant characteristics at enrollment (like age, gender, duration of smoking, nicotine dependence, and CO reading) were compared between the three smoking-form subgroups using formal statistical tests (for mean: Tukey’s least significant difference test; for median: Kruskal–Wallis test; and for nominal: chi-square test). Subgroup-specific relative risks (RRs) of smoking abstinence and CIs were computed and tested for differential intervention effect among the three forms of smoking using homogeneity test statistics (likelihood ratio estimates for Type 3 analysis) (Assmann, Pocock, Enos, & Kasten, 2000; Lagakos, 2006; Wang, Lagakos, Ware, Hunter, & Drazen, 2007) in SAS, version 9.3 (SAS Institute) using Proc Genmod (generalized estimating equations models). Estimates adjusted for the covariates—age, gender, and duration of smoking—are also presented. These adjusted estimates (adj. RRs) are used to construct a forest plot, with a log RR scale and marker size corresponding to sample weight in each subgroup, for further visual interpretation. It is very important to realize that the effect size estimates and CIs in these analyses are not presented to show statistical significance; rather, they are interpreted as providing a plausible range of intervention differences, consistent overall with the main trial results (Lagakos, 2006).
Results The primary outcome data were missing for 5.8% of the total sample, that is, 15 hookah smokers, 74 cigarette smokers, and 25 mixed smokers. There was no differential loss to followup or intervention completion among the three subgroups (Figure 1). There were significant differences among the three subgroups at enrollment: hookah-only smokers were older—with
683
Behavioral support for hookah smokers Trial participants in total: 1955
Subgroups of smoking form
Hookah BS+: 27 BS: 118 Control: 70 Total: 215
Cigarette
Mixed
BS+: 465 BS: 371 Control: 419 Total: 1255
BS+: 167 BS: 151 Control: 167 Total: 485
Did not complete intervention BS+: 4 BS: 1
BS+: 28 BS: 6
BS+: 8 BS: 1
Loss to follow-up BS+: 7 BS: 2 Control: 6 Total: 15
BS+: 37 BS: 13 Control: 24 Total: 74
BS+: 9 BS: 5 Control: 11 Total: 25
Analyzed BS+: 20 BS: 116 Control: 64 Total: 200
BS+: 428 BS: 358 Control: 395 Total: 1181
BS+: 158 BS: 146 Control: 156 Total: 460
Figure 1. Flow chart of subgroups of smoking forms enrolled. more females, higher CO readings, lower nicotine dependency score, and longer duration of smoking—than the cigarette-only smokers (Table 1). The test for homogeneity of intervention effects among these three subgroups was statistically significant (p-value for BSS+: .04 and that for BSS: .02), both for the crude and the adjusted analyses (Table 2). Compared to the control, the RR of smoking abstinence for BSS+ was 2.5 (95% CI = 1.3–4.7) and that for BSS was 2.2 (95% CI = 1.3–3.8) among hookah-only smokers; and the RR for BSS+ was 6.6 (95% CI = 4.6–9.6) and that for BSS was 5.8 (95% CI = 4.0–8.5) among cigarette-only smokers(Table 2). After adjustment for age, gender, and duration of smoking, the effect estimates essentially remained the same (Table 2).
Discussion These findings suggest that behavioral support with and without bupropion appears to be effective in achieving 6-month
684
smoking abstinence among hookah-only smokers but to a lesser extent than among cigarette-only smokers. It is possible that the differences observed among the three subgroups might be (a) due to the variation in characteristics of the individuals, the way they smoke (habits regarding puff depth and frequency), and sociocultural determinants or (b) due to differences in nicotine dependency levels. The BCTs used in our intervention were primarily developed to influence behavioral determinants of cigarette smoking. However, unique behaviors and stimuli connected to the nicotine delivery products are likely to differ across the various forms of smoking, particularly hookah smoking. Several potential limitations of this post-hoc analysis should be considered. Tests of interaction typically have low statistical power, and the sample size calculation for this trial did not take into account the power to detect heterogeneity of intervention effects by smoking forms. However, there were sufficiently large numbers within each subgroup to allow adequate analysis. Risk of false-positive results increases with multiple subgroup testing; therefore, the analysis was restricted to the
Nicotine & Tobacco Research Table 1. Participant Characteristics by Subgroups of Smoking Form at Enrollment: ASSIST Pakistan Trial Characteristic Mean age (SD) in years Number of females (%) Median household income per month (IQR) in USD Median duration of smoking (IQR) in years Median smoking start age in years (IQR) Median quantity smoked daily (IQR)a Sample (interventions only) Mean CO reading at recruitment (SD) Mean Fagerström score (SD) Number (%) of patients by severity of nicotine dependency Very high High Medium Low Very low
Cigarette smokers (n = 1,255)
Mixed smokers (n = 485)
Hookah smokers (n = 215)
p value
37.4 (11.5) 21 (1.7) 81.4 (58.1)
45.4 (13.3) 24 (5.0) 93.0 (81.4)
51.5 (13.8) 45 (21.0) 81.4 (69.8)
