Abstract
Introduction
A nicotine product standard reducing the nicotine content in cigarettes could improve public health by reducing smoking. This study evaluated the potential unintended consequences of a reduced nicotine product standard by examining its effects on (1) smoking behaviors based on drinking history; (2) drinking behavior; and (3) daily associations between smoking and drinking.
Methods
Adults who smoke daily (n = 752) in the United States were randomly assigned to smoke very low nicotine content (VLNC) cigarettes versus normal nicotine content (NNC; control) cigarettes for 20 weeks. Linear mixed models determined if baseline drinking moderated the effects of VLNC versus NNC cigarettes on Week 20 smoking outcomes. Time-varying effect models estimated the daily association between smoking VLNC cigarettes and drinking outcomes.
Results
Higher baseline alcohol use (vs no use or lower use) was associated with a smaller effect of VLNC on Week 20 urinary total nicotine equivalents (ps < .05). No additional moderation was supported (ps > .05). In the subsample who drank (n = 415), in the VLNC versus NNC condition, daily alcohol use was significantly reduced from Weeks 17 to 20 and odds of binge drinking were significantly reduced from Weeks 9 to 17. By Week 7, in the VLNC cigarette condition (n = 272), smoking no longer predicted alcohol use but remained associated with binge drinking.
Conclusions
We did not support negative unintended consequences of a nicotine product standard. Nicotine reduction in cigarettes generally affected smoking behavior for individuals who do not drink or drink light-to-moderate amounts in similar ways. Extended VLNC cigarette use may improve public health by reducing drinking behavior.
Implications
There was no evidence that a VLNC product standard would result in unintended consequences based on drinking history or when considering alcohol outcomes. Specifically, we found that a very low nicotine standard in cigarettes generally reduces smoking outcomes for those who do not drink and those who drink light-to-moderate amounts. Furthermore, an added public health benefit of a very low nicotine standard for cigarettes could be a reduction in alcohol use and binge drinking over time. Finally, smoking VLNC cigarettes may result in a decoupling of the daily associations between smoking and drinking.
In 2018, the FDA issued an Advanced Notice of Proposed Rulemaking announcing their intent to pursue a nicotine reduction policy for cigarettes. A policy mandating a reduction in the nicotine content of cigarettes could reduce cigarette smoking and the associated 480 000 US annual deaths1 by reducing the addictiveness of cigarettes.2 Several randomized clinical trials (RCTs) indicate that very low nicotine content (VLNC) cigarettes (0.04 mg nicotine/g tobacco) reduce the number of cigarettes per day (CPD), toxicant exposure, and nicotine dependence, and can facilitate quit attempts when compared with normal nicotine content (NNC; 15.5 mg nicotine/g tobacco) cigarettes.2–8
Although a nicotine reduction policy is expected to improve public health by reducing smoking, researchers have evaluated how this policy would affect vulnerable subgroups with co-occurring psychiatric diagnoses9,10 or substance use.11,12 One such subgroup includes individuals who exhibit heavy or problematic drinking patterns, who have high rates of smoking, and who have low rates of cessation.13–15 The elevated risk for these individuals may be partly explained by direct causal pathways linking alcohol use with smoking.16–19 Therefore, it is important to establish whether one’s drinking history alters the effectiveness of a nicotine reduction policy. Previously, a multisite RCT supported that 6 weeks of VLNC cigarettes (vs NNC cigarettes) significantly reduced CPD and nicotine exposure.2 Subsequent analyses of that trial found that the intervention effects were not moderated by baseline alcohol use or problem drinking.12 Additional research is needed to determine whether history of alcohol use affects response to the policy over longer periods of VLNC cigarette exposure.
In addition to considering the effectiveness of the intervention among individuals who drink, it is important to understand the impact of the intervention on alcohol consumption. Alcohol use is another leading preventable contributor to morbidity and death in the United States,20 and based on laboratory experiments, nicotine exposure increases alcohol use.16,19 Therefore, reducing nicotine exposure via a nicotine reduction policy could decrease alcohol use,17 which would further benefit public health. Analyses of the aforementioned 6-week clinical trial did not support a direct effect of VLNC cigarette use on alcohol outcomes; however, indirect/mediational pathways were supported.12 Specifically, VLNC cigarette use was associated with significant reductions in nicotine exposure and CPD, which in turn, corresponded with reductions in alcohol use. The indirect pathways linking the intervention to reduced drinking were significant from Weeks 2 to 6, but not from Weeks 0 to 2, of the intervention. These findings suggest that VLNC cigarette use could decrease alcohol outcomes, but this effect may emerge over time.
It is not surprising that the effects of nicotine reduction on related outcomes, like alcohol use, would be dynamic.17 The effects of nicotine reduction on nicotine exposure and smoking behavior are asynchronous. Switching to VLNC cigarettes leads to an immediate and substantial reduction in nicotine exposure,21–23 whereas changes in smoking appeared to take several weeks to emerge and then continuing to subsequently decrease.4,5,22 Potentially, both the immediate changes in nicotine exposure and the more gradual changes in smoking behavior may have independent effects on alcohol outcomes that are dynamic over time.
Given the potential for dynamic effects of VLNC cigarette use on drinking, it is important to evaluate such effects over extended time periods with time-sensitive analyses. The time-varying effect model (TVEM) can identify dynamic intervention effects.24 Unlike conventional RCT analytic methods, which focus on a single end point or average effects across multiple repeated measurements, TVEM can capture temporal changes in the relation between the intervention and the outcome variables of interest. Therefore, TVEM can help identify at what time periods VLNC cigarette use affects alcohol consumption.
To better understand the interplay of alcohol use and a nicotine reduction policy, this study aimed to extend prior work by examining effects over a longer period of time (20 vs 6 weeks) and using TVEM (vs structural equation modeling) to evaluate how these effects emerge over time. The aims of this study include (1) evaluating if baseline drinking characteristics moderate the effectiveness of 20 weeks of VLNC cigarette use on smoking-related outcomes; (2) using TVEM during this 20-week intervention to examine how VLNC cigarette use affects drinking behavior over time; and (3) exploring if the covariation of daily smoking and drinking behavior changes over time as a result of switching to VLNC cigarettes. Prior analysis of this data found that being assigned to smoke VLNC cigarettes for 20 weeks (vs NNC cigarettes) significantly reduced CPD, carbon monoxide (CO), and creatinine-adjusted total nicotine equivalents (TNE).4 We hypothesized that baseline drinking characteristics would not alter/moderate the previously supported effectiveness of VLNC cigarette use on smoking-related outcomes. Furthermore, we predicted that VLNC cigarette use would result in a reduction in alcohol outcomes but that this effect would increase over time. Finally, we hypothesized that the association between daily smoking and drinking would become “decoupled” over time in the VLNC cigarette condition. That is, smoking would become less associated with drinking over time, consistent with a weakening of the cue–response relationship due to repeated nicotine reduction.
Methods
Participants
Adults (N = 1250) were recruited from 10 sites including Duke University, Johns Hopkins University, Mayo Clinic, MD Anderson Cancer Center, Oregon Research Institute, University of California San Francisco, University of Minnesota–Twin Cities, University of Minnesota-Duluth, University of Pennsylvania, and Moffitt Cancer Center. Participants were recruited by each site through advertisements such as television, radio, internet, direct mailing, and flyers. Eligibility criteria at the screening visit included being ≥18 years old, smoking ≥ five CPD, and expired breath CO level > 8 ppm (or urinary cotinine of >1000 ng/mL). Alcohol-related exclusion criteria included breath alcohol concentration > 0.01 (allowed to re-screen once) and binge drinking alcohol (four/five drinks per day for female/male) > 9 days in the past 30 days. Additional study exclusion criteria included intention to quit smoking in the next 30 days, use of other tobacco products on more than 9 days out of past 30, exclusive use of roll-your-own cigarettes, prior exposure to reduced-nicotine-content cigarettes, serious or unstable psychiatric or medical illness, positive urine screen for illicit drugs other than cannabis, and breastfeeding, pregnancy, or planning to become pregnant.
Study Design and Procedures
This is a secondary analysis of data from a double-blind RCT (see main paper for details, like power justifications, briefly summarized here).4 Although the original study was preregistered (ClinicalTrials.gov Identifier: NCT02139930), this secondary data analysis was not. After a 2-week baseline usual brand smoking phase, participants were randomly assigned to receive Spectrum research cigarettes,25 with either immediate nicotine reduction, gradual nicotine reduction, or NNC cigarettes (control condition), respectively. This article focused on all available data from the relevant subset of participants (N = 752) who were assigned to either the immediate nicotine reduction (VLNC; n = 503) or NNC (n = 249) conditions. For 20 weeks, participants in the VLNC condition received research cigarettes with 0.4 mg nicotine/g tobacco, whereas participants in the NNC condition received research cigarettes with 15.5 mg nicotine/g tobacco. Participants received menthol or non-menthol cigarettes matched to their preference. As described in more detail in the primary paper,4 participants were asked not to smoke any cigarettes other than their assigned study cigarettes, which was incentivized with bonus payments. Participants attended weekly laboratory visits for the first 4 weeks after randomization and then biweekly for the next 16 weeks. This secondary analysis was ruled exempt from review by the Ryerson University ethics board. The original data collection was approved by each site’s ethics board.
Measures
Nicotine Dependence
Participants completed the Fagerström Test for Nicotine Dependence (FTND).26
Alcohol Problems
History of drinking problems were assessed at the baseline visit using the Michigan Alcohol Screening Test—Short Form (SMAST),27 which is scored as the sum of responses to 13 yes/no questions about drinking difficulties, with >4 indicating problem drinking.
Alcohol Outcomes
A daily record of standard drinks consumed during the baseline period or since the last experimental visit were assessed using a timeline follow-back (TLFB) interview.28 Daily binge drinking (defined as consuming ≥4 drinks for women or ≥5 drinks for men within 2 hours) was also reported.
Cigarette Outcomes
Throughout the study, participants completed daily phone calls assessing use of study and nonstudy CPD via an automated interactive voice response system. Study cigarettes were those provided to the participants, and nonstudy cigarettes were other cigarettes not provided (eg, usual brand of cigarette).
Biomarkers
Breath CO and urine TNE (first void) were measured to determine nicotine exposure. TNE adjusted for creatinine was computed as the sum of nicotine and six metabolites: total nicotine, total cotinine, total trans 3′-hydroxycotinine (sum of the analyte and respective glucuronide conjugate), and nicotine-N-oxide.
Analyses
Baseline Drinking Characteristics as a Moderator
Linear mixed models were conducted using R version 3.4.2.29 Analyses included participants in the VLNC and NNC conditions (n = 752). The dependent variables were total CPD (study plus nonstudy CPD), CO, and TNE at Weeks 4, 8, 12, 16, and 20. Each model included a random intercept to account for repeated measures from the same individual, and fixed effects for intervention condition, a moderator, intervention and moderator interaction, visit, visit and moderator interaction, visit and intervention interaction, and visit, moderator and intervention interaction, as well as covariate adjustments mentioned below. The focus of our analysis was the treatment effect at week 20, which was estimated using the appropriate contrast from the linear mixed model. The moderators included baseline drinks per day from the TLFB and SMAST scores. The moderators were examined in separate models as continuous and categorical variables (ie, with 1 indicating any alcohol use or problem SMAST score). For both the continuous and categorical versions of baseline drinking and SMAST, three models were performed: (1) adjusted for baseline value of the outcome; (2) adjusted for baseline value of the outcome, study site, and baseline variables that differed between treatment arms with p < .2 (employment status, FTND, and nicotine metabolite ratio; adjusted for in the primary paper); and (3) adjusted for the same covariates as #2 with the addition of variables associated with the moderator at baseline (age, gender, and race for baseline drinking and age, gender, race, and ethnicity for SMAST). Additionally, each model was adjusted for baseline variables that differed between week 20 completers and noncompleters with p < .2 (age, gender, ethnicity, TNE, FTND, menthol status, and baseline drinking) if not already included in the model. Other tobacco products and years of regular smoking also differed but were excluded from the analysis due to significant missing data (other tobacco products) or correlation with age (years of regular smoking). The linear mixed model adjusted for covariates associated with missing data will be valid assuming that the data are missing at random.30 TNE was highly skewed and was modeled on the log-scale and back-transformed as a ratio of geometric means.31,32
Time-Varying Effects of Intervention on Daily Alcohol Outcomes
TVEM was completed using SAS v9.4 using %TVEM macro v3.11.33 These analyses focused on a subsample (n = 415) of individuals in the VLNC (n = 272) or NNC (n = 143) conditions who reported recent alcohol use (ie, any alcohol use in the past month at screening or on the TLFB during baseline), had data on applicable covariates, and had at least one response on the alcohol TLFB during the experimental period. There were 57 492 possible observations for this subsample, and the proportion of missing observations is described with each analysis alongside the results.
TVEM examined the time-varying effect of intervention (VLNC vs NNC) on daily alcohol use for the 20-week period (ie, Day 1 up to Day 140). The Poisson distribution was specified for alcohol use; therefore, daily alcohol use was rounded to the nearest whole number to align with a count outcome. Daily alcohol outcomes were modeled using penalized splines (P-spline method with 10 knots), which automatically select the proper number of knots (ie, amount of smoothness).33 Robust sandwich estimators were used to account for within-subject correlation. We relied on quadratic splines, as opposed to cubic splines, because they are more parsimonious and complex functions were not anticipated. Missing observations were excluded from the analyses (ie, pairwise deletion at the level of observation). The analyses were repeated with the binary binge drinking outcome variable utilizing logistic TVEM. Using product terms, gender and SMAST scores were examined as moderators of the intervention effect. Analyses controlled for study site, employment status, FTND, nicotine metabolite ratio, age, gender, and Black or other race.
Time-Varying Associations Between Daily Cigarette and Alcohol Use
Using TVEM, the time-varying association between CPD and alcohol use over time was examined in the subsample of alcohol users using two approaches. First, daily study cigarette use was examined as a time-varying predictor of daily alcohol use and binge drinking. A product term between intervention and daily study cigarette use was included to examine its time-varying effect. Second, using just the VLNC subsample of individuals who drink alcohol (n = 272), we separately examined the time-varying associations of daily study cigarette use or nonstudy cigarette use with daily alcohol use or binge drinking. These analyses were accomplished with the same TVEM approach and covariates described above.
Results
Baseline Sample Descriptives
The sample mean age (n = 752) was 45.36 (SD = 13.70). It was 45.08% female, 62.52% White, 29.36% Black, 8.12% other race, and 5.59% Hispanic. The mean SMAST score was 3.01 (SD = 2.13, range: 0–11), and FTND was 5.36 (SD = 2.12, observed range: 0–10). The sample averaged 17.25 CPD (SD = 8.60). 53.46% reported alcohol use and 17.29% had scores > 4 on the SMAST. The reported mean number of standard drinks per day was 0.57 (SD = 0.98), and mean percentage of binge drinking days was 1.02 (SD = 3.94).
Among participants who reported past-month alcohol use at baseline or any alcohol use during the baseline period (n = 415), the mean age was 43.74 (SD = 14.37), and it was 44.58% female, 65.78% White, 26.75% Black, 7.47% other race, and 3.86% Hispanic. The mean SMAST was 3.32 (SD = 1.67), and 8.67% had SMAST scores > 4 (which indicates having alcohol problems). Mean FTND was 4.98 (SD = 2.09). On average, the subsample reported drinking 0.93 standard drinks per day (SD = 1.08), and mean percentage of binge drinking days was 1.64% (SD = 4.06).
Baseline Drinking Characteristics as a Moderator of Intervention Effects on Smoking Outcomes
The results of the moderation analyses are summarized in Table 1. At Week 20, there were 179 individuals with missing data on CPD, 197 on CO, and 198 on creatinine-corrected TNE. Subgroup analyses in Table 1 demonstrate that all the drinking subgroups exhibited significant intervention effects for the smoking outcomes. Furthermore, baseline drinking did not moderate the effect of VLNC versus NNC condition on Week 20 CPD or CO. These results were consistent in both unadjusted and adjusted regression models and when examining baseline drinking continuously and dichotomously. Baseline drinking significantly moderated the effect of VLNC on TNE in the unadjusted and adjusted models (ps < .05). Specifically, the intervention led to a greater reduction in TNE for those who reported no alcohol use at baseline relative to those who reported baseline alcohol use. In all of the models, baseline SMAST score was not a significant moderator of the effect of VLNC versus NNC condition on Week 20 CPD, CO, and creatinine-adjusted TNE.
Table 1.
Effect of Baseline Drinking as a Moderator on Week 20 Outcomes Using Linear Mixed-Effect Model Method
Baseline drinking as moderator | |||||
---|---|---|---|---|---|
Treatment effect (95% CI) | Treatment effect (95% CI) | ||||
Outcome | Modela | Baseline drinks = 0 (N = 350) | Baseline drinks > 0 (N = 402) | p-Value for categorical moderator interaction | p-Value for continuous moderator interaction |
CPD | Unadjustedb | −8.23 (−9.99, −6.46) | −8.65 (−10.30, −7.00) | .74 | .15 |
Adjustedc | −8.02 (−9.80, −6.23) | −8.69 (−10.33, −7.04) | .59 | .14 | |
Adjustedd | −8.11 (−9.91, −6.30) | −8.75 (−10.42, −7.07) | .62 | .18 | |
Creatinine-corrected TNE | Unadjustedb | 0.25 (0.18, 0.37) | 0.14 (0.10, 0.20) | .03 | .0498 |
Adjustedc | 0.27 (0.18, 0.39) | 0.14 (0.10, 0.20) | .02 | .05 | |
Adjustedd | 0.28 (0.19, 0.40) | 0.14 (0.10, 0.20) | .01 | .046 | |
CO | Unadjustedb | −5.09 (−7.09, −3.10) | −6 (−7.87, −4.13) | .52 | .23 |
Adjustedc | −4.93 (−6.96, −2.90) | −6.16 (−8.03, −4.28) | .39 | .19 | |
Adjustedd | −4.78 (−6.83, −2.75) | −6.12 (−8.02, −4.23) | .35 | .18 | |
Baseline SMAST as moderator | |||||
Treatment effect (95% CI) | Treatment effect (95% CI) | ||||
Outcome | Modela | Baseline SMAST ≤ 4 (N = 622) | Baseline SMAST > 4 (N = 130) | p-Value for categorical moderator interaction | p-Value for continuous moderator interaction |
CPD | Unadjustedb | −8.03 (−9.37, −6.70) | −10.35 (−13.14, −7.56) | .14 | .91 |
Adjustedc | −8.11 (−9.45, −6.76) | −9.63 (−12.42, −6.83) | .34 | .61 | |
Adjustedd | −8.22 (−9.59, −6.85) | −9.58 (−12.4, −6.74) | .40 | .52 | |
Creatinine-corrected TNE | Unadjustedb | 0.20 (0.15, 0.27) | 0.12 (0.07, 0.22) | .14 | .08 |
Adjustedc | 0.20 (0.15, 0.27) | 0.14 (0.08, 0.25) | .29 | .13 | |
Adjustedd | 0.20 (0.15, 0.27) | 0.15 (0.08, 0.26) | .33 | .14 | |
CO | Unadjustedb | −5.77 (−7.29, −4.25) | −4.76 (−7.92, −1.6) | .57 | .90 |
Adjustedc | −5.79 (−7.32, −4.26) | −4.66 (−7.84, −1.48) | .54 | .93 | |
Adjustedd | −5.71 (−7.26, −4.16) | −4.57 (−7.77, −1.38) | .54 | .85 |
The table summarizes the results for the unadjusted and adjusted moderation models for the following outcomes: cigarettes per day (CPD), creatinine-corrected total nicotine equivalents (TNE), and carbon monoxide (CO) at week 20. The top panel summarizes the results for analyses evaluating baseline drinking as a moderator. Specifically, we report the Treatment Effects (the regression coefficients and 95% CI) separately for nondrinking and drinking subsample. Then, in the last two columns, we provide the p-values for the moderation analyses when examining baseline drinking as a categorical moderator and continuous moderator, respectively. The bottom panel summarizes the same information but with baseline Michigan Alcohol Screening Test—Short Form (SMAST) as the moderator.
aAll models are adjusted for variables that were different between completers and noncompleters (age, gender, ethnicity, log-transformed creatinine-corrected TNE, Fagerström Test for Nicotine Dependence, menthol status, and baseline drinking) if not already included in the model. Other tobacco products and years of regular smoking also differed but were excluded from the analysis due to significant missing data (other tobacco products) or correlation with age (years of regular smoking).
bAdjusted for baseline outcome.
CAdjusted for baseline outcome, site, and baseline variables that were different between treatment arms (employment, Fagerström Test for Nicotine Dependence, and serum nicotine metabolic ratio). There were 17 subjects with missing cotinine data who were not included in this analyses.
dAdjusted for baseline outcome, site, baseline variables that were different between treatment arms (employment, Fagerström Test for Nicotine Dependence, and serum nicotine metabolic ratio), and baseline variables that were associated with the moderator (age, race, and gender). The analyses with the SMAST moderator also included ethnicity as a covariate. There were 17 subjects with missing cotinine data and 13 with missing race data who were not included in this analyses.
Time-Varying Effects of Intervention on Daily Alcohol Outcomes
Alcohol Use
There were 48 787 complete observations and 8705 missing observations (15.14% missing of the aforementioned 57 472 total) across the 415 participants. There was a time-varying effect of VLNC versus NNC cigarettes on daily alcohol use (Figure 1a). Specifically, the effect of VLNC cigarette use on alcohol use emerged gradually over time such that daily alcohol use became significantly reduced in the VLNC group relative to the NNC group starting on Day 125 (Week 17). This effect continued to strengthen slightly until the end of the observation period (Day 140). The effect of cigarette assignment on daily alcohol use was not moderated by SMAST score (estimate = −0.03, 95% confidence interval [CI]: −0.17 to 0.11, p = .68) or gender (estimate = −0.19, 95% CI: −0.66 to 0.28, p = .43).
Figure 1.
The time-varying effect of immediate reduction of nicotine in cigarettes (versus control) on daily alcohol outcomes: (a) The time-varying Poisson coefficient (solid line) and corresponding 95% confidence interval (dashed lines) for daily alcohol use. (b) The time-varying logistic regression coefficient (solid line) and corresponding 95% confidence interval (dashed lines) for daily alcohol use. For (a) and (b), Day (x-axis) is the day in the trial, and Coefficient (y-axis) is the estimated coefficient for the analysis. Areas where the 95% confidence interval does not contain 0 (marked by the horizontal axis line) are shaded with vertical lines and support that during those time periods there are significant differences in the outcome (daily alcohol use or binge drinking) between the immediate nicotine reduction and control conditions.
Binge Drinking
There was a total of 48 039 complete and 9453 (16.44%) missing observations for the 415 participants. There was also a time-varying effect of VLNC versus NNC cigarettes on daily binge drinking (Figure 1b). Specifically, the effect of cigarette assignment on the log odds of binge drinking increased over time, such that by Day 72, binge drinking was significantly reduced in the VLNC group relative to the NNC group. This effect continued to increase until Day 98 and then began to diminish such that there was no longer a significant intervention effect after Day 127. The effect of intervention on daily binge drinking was not moderated by SMAST score (estimate = −0.13, 95% CI: −0.38 to 0.12, p = .30) or gender (estimate = 0.12, 95% CI: −0.86 to 1.10, p = .81).
Time-Varying Associations Between Daily Cigarette and Alcohol Use
RCT Condition as a Moderator
For the 415 participants using alcohol, there were 45 261 complete observations (12 231 [21.27%] missing) for the alcohol use and 44 567 (12 925 [22.48%] missing) for binge drinking outcome analysis. Controlling for the covariates, the association between daily study CPD and daily alcohol use was not moderated by cigarette condition (estimate = −0.01, 95% CI: −0.03, 0.01, p = .19 for time-invarying interaction effect). Similarly, cigarette condition did not moderate the association between daily study CPD and daily binge drinking (estimate = 0.02, 95% CI: 0.0004 to 0.04, p = .21 for time-invarying interaction effect).
VLNC Condition Only
For the 272 participants using alcohol in the VLNC condition, there were 37 686 possible observations. There were 28 193 complete observations (25.19% missing) for alcohol use and 27 645 (26.64% missing) for binge drinking. Controlling for covariates, daily study (ie, VLNC) CPD was not significantly associated with daily alcohol use for the entirety of the study (Figure 2). In contrast, there was a time-varying association between daily nonstudy CPD and daily alcohol use. Specifically, there was initially a positive association between nonstudy CPD with daily alcohol use, but the magnitude of this relation decreased over time such that it was no longer significant by Day 43 (Week 7). Study CPD was positively and significantly associated with the occurrence of binge drinking from Day 15 to 52 only (Figure 3). In contrast, nonstudy CPD had a positive and significant association with the occurrence of binge drinking throughout the observation period, except from Day 54 to 91, the association was nonsignificant.
Figure 2.
Time-varying associations between daily non-study and study cigarette use with daily alcohol use in the immediate reduction group: The figure depicts the time-varying relationship between daily nonstudy cigarettes per day (CPD) with daily alcohol use (in black) and daily study CPD with daily alcohol use (in gray) within the immediate reduction group only. The solid lines show the Poisson coefficient estimate (y-axis) across the days of the study period (x-axis), and the dashed lines correspond with the 95% confidence interval. The area shaded with vertical lines summarizes where the 95% confidence interval does not contain 0 (marked with the horizontal axis line from 0), and thus is the period of time when the coefficient is statistically significant.
Figure 3.
Time-varying associations between daily non-study and study cigarette use with daily binge drinking in the immediate reduction group: The figure depicts the time-varying logistic regression coefficients between daily nonstudy cigarettes per day (CPD) with daily binge drinking (in black) and daily study CPD with daily binge drinking (in gray) within the immediate reduction group only. The solid lines show the logistic regression coefficient estimate (y-axis) across the days of the study period (x-axis), and the dashed lines correspond with the 95% confidence interval. The shaded areas with vertical lines (black for nonstudy CPD and gray for study CPD) summarize where the 95% confidence interval does not contain 0 (marked with the horizontal axis line from 0), and thus is the period of time when the coefficient is statistically significant.
Discussion
Implementing a policy that reduces the nicotine content of cigarettes to very low levels could greatly reduce the negative public health effects of smoking. We found no evidence that a very low nicotine product standard would result in unintended consequences (1) on smoking outcomes based on drinking history or (2) on alcohol outcomes. Baseline drinking moderated intervention effects on TNE at Week 20. Specifically, VLNC cigarettes reduced TNE at Week 20 for participants who did and did not drink, but the reduction was relatively larger for those who did not drink alcohol. This may have occurred for a number of unexamined reasons. Alcohol consumption could reduce adherence with smoking VLNC cigarettes exclusively (particularly when intoxicated), resulting in a relatively smaller reduction in nicotine biomarkers over time. Consistent with prior research,12 moderation was not supported for the other smoking outcomes or moderators; therefore, it is recommended that the moderation effect with TNE be interpreted with caution. Taken together with the results of the primary paper, a nicotine reduction policy would be effective in reducing smoking-related outcomes4 for individuals with varied drinking histories including those who do or do not drink and those with alcohol-related problems.
Evidence also supported that a nicotine reduction policy could have additional public health benefit by reducing alcohol use. We replicated the prior finding of no direct effect of VLNC cigarette use on alcohol outcomes during the first 6 weeks.12 With the longer follow-up period and new analytic strategy, however, we found that switching to VLNC cigarettes eventually significantly reduced alcohol use. The reduction in daily alcohol use occurred after 17 weeks of VLNC cigarette use and persisted through the end of the 20-week trial. There was also evidence of reduced binge drinking in response to VLNC cigarette use; however, this was specific to Weeks 9–17. It is not clear why the reduction occurred only during this time period, and replication should be attempted. More broadly, this pattern of findings emphasizes that it is crucial to evaluate the long-term impact of policies like a nicotine reduction standard as their impact on some health outcomes can emerge gradually over time.
From a mechanistic standpoint, it is interesting to consider why it would take over 2 months of VLNC cigarette use for alcohol use to change. Given that nicotine exposure was immediately and substantially reduced,4 other processes besides direct nicotine and alcohol pharmacological interactions may be contributing to this finding. One possible explanation is that repeated use of VLNC cigarettes changed the extent to which smoking behavior prompts alcohol use. For example, individuals who were assigned to VLNC cigarettes exhibited a decoupling of the relation between the use of nicotine-containing cigarettes and alcohol over time. Although it is premature to conclude why this decoupling occurred, we suspect that repeated VLNC cigarette use interrupts the cue–response relationship between cigarette smoking and alcohol use.17 It is also important to acknowledge that for the duration of the 20-week trial, the number of VLNC cigarettes smoked was unrelated to level of alcohol use. This suggests that the very low levels of nicotine in cigarettes did not prompt alcohol use, as a usual brand cigarette would. Alternatively, given the positive association between nonstudy cigarette and alcohol use described above, it suggests that individuals preferred to smoke their usual brand of cigarette on days that they drink. It is possible that this pattern of nonadherence would delay the decoupling of smoking and alcohol use.
Unlike the results with alcohol use, binge drinking was no longer significantly reduced in the VLNC cigarette condition relative to control condition at the end of the study. Furthermore, nonstudy CPD had a positive and significant association with binge drinking throughout most of the study, except for Weeks 8–13 when the confidence intervals were wider and the coefficients were slightly decreased. Therefore, a stable association may underlie these findings as the relation largely persisted. It is unclear why the reduction in binge drinking was no longer significant at the end of the study. This pattern of findings does suggest that VLNC cigarette use may be more effective in reducing alcohol use in general, as opposed to more problematic drinking behavior. This difference could, in part, be explained by nonadherence behavior specifically when binge drinking (as opposed to drinking smaller quantities).
There are limitations to consider when interpreting these results. Data were collected on daily substance use, as opposed to moment-to-moment use within a day, which prevented us from examining how co-use relationships fluctuated within substance-using episodes. Furthermore, participants used nonstudy cigarettes, particularly in the VLNC group, which could affect nicotine exposure and undermine the effectiveness of the intervention, and also likely mitigated any adverse effects. Finally, the generalizability of the findings to the general population of individuals who smoke and drink is unknown. There were exclusionary criteria related to heavy drinking for this sample; however, relatively few people were excluded as a result. Specifically, at screening, 37 participants were excluded due to drinking-related criteria. An additional eight individuals dropped out of the study following positive breath alcohol concentration readings (that required rescheduling the appointment). Furthermore, drinking severity was relatively low among the drinking subgroup. It is important for future research to evaluate response to this intervention in samples who drink heavily or are alcohol dependent.
This study provides further evidence that, irrespective of non-heavy drinking patterns, a very low nicotine standard for cigarettes can improve public health by significantly decreasing cigarette use. We did not observe any unintended negative consequences of a nicotine reduction in cigarettes on smoking outcomes based on one’s drinking history. Furthermore, an added public health benefit could be reduction in alcohol use and possibly binge drinking, and these reductions could emerge gradually over time. It appears that a reduced nicotine standard in cigarettes also has the potential to alter co-use patterns, by decoupling the relation between smoking and drinking. As the co-use of alcohol and cigarettes are associated with much greater health harms than using either substance alone,34 reducing co-use could further improve public health.
Supplementary Material
A Contributorship Form detailing each author’s specific involvement with this content, as well as any supplementary data, are available online at https://academic.oup.com/ntr.
Funding
Research supported by a grant from the National Institute on Drug Abuse (NIDA) and the Food and Drug Administration (FDA) Center for Tobacco Products (U54 DA031659). Research reported in this publication was supported by National Institutes of Health (NIH) grant P30 CA77598 utilizing the Biostatistics and Bioinformatics Core shared resource of the Masonic Cancer Center, University of Minnesota and by the National Center for Advancing Translational Sciences of the National Institutes of Health Award Number UL1TR0002494. Author support also included NIDA T32 DA016184. The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the NIH or the FDA.
Declaration of Interests
DH, JSK, ED, and JJ reported receiving grants from the National Institute on Drug Abuse. DJD reported serving as a paid expert witness in litigation against tobacco companies. No other conflict of interests apply.
References
- 1. US Department of Health Human Services. The Health Consequences of Smoking—50 Years of Progress: A Report of the Surgeon General. Atlanta, GA: US Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health; 2014. [Google Scholar]
- 2. Donny EC, Denlinger RL, Tidey JW, et al. Randomized trial of reduced-nicotine standards for cigarettes. N Engl J Med. 2015;373(14):1340–1349. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Benowitz NL, Hall SM, Stewart S, Wilson M, Dempsey D, Jacob P 3rd. Nicotine and carcinogen exposure with smoking of progressively reduced nicotine content cigarette. Cancer Epidemiol Biomarkers Prev. 2007;16(11):2479–2485. [DOI] [PubMed] [Google Scholar]
- 4. Hatsukami DK, Luo X, Jensen JA, et al. Effect of immediate vs gradual reduction in nicotine content of cigarettes on biomarkers of smoke exposure: a randomized clinical trial. JAMA. 2018;320(9):880–891. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Hatsukami DK, Kotlyar M, Hertsgaard LA, et al. Reduced nicotine content cigarettes: effects on toxicant exposure, dependence and cessation. Addiction. 2010;105(2):343–355. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Donny EC, Houtsmuller E, Stitzer ML. Smoking in the absence of nicotine: behavioral, subjective and physiological effects over 11 days. Addiction. 2007;102(2):324–334. [DOI] [PubMed] [Google Scholar]
- 7. Shiffman S, Kurland BF, Scholl SM, Mao JM. Nondaily smokers’ changes in cigarette consumption with very low-nicotine-content cigarettes: a randomized double-blind clinical trial. JAMA Psychiatry. 2018;75(10):995–1002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Smith TT, Hatsukami DK, Benowitz NL, et al. Whether to push or pull? Nicotine reduction and non-combusted alternatives – two strategies for reducing smoking and improving public health. Prev Med. 2018;117:8–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Tidey JW, Pacek LR, Koopmeiners JS, et al. Effects of 6-week use of reduced-nicotine content cigarettes in smokers with and without elevated depressive symptoms. Nicotine Tob Res. 2017;19(1):59–67. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Tidey JW, Colby SM, Denlinger-Apte RL, et al. Effects of 6-week use of very low nicotine content cigarettes in smokers with serious mental illness. Nicotine Tob Res. 2019;21(suppl 1):S38–S45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Pacek LR, Vandrey R, Dermody SS, et al. Evaluation of a reduced nicotine product standard: moderating effects of and impact on cannabis use. Drug Alcohol Depend. 2016;167:228–232. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Dermody SS, Tidey JW, Denlinger RL, et al. The impact of smoking very low nicotine content cigarettes on alcohol use. Alcohol Clin Exp Res. 2016;40(3):606–615. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Weinberger AH, Gbedemah M, Goodwin RD. Cigarette smoking quit rates among adults with and without alcohol use disorders and heavy alcohol use, 2002–2015: a representative sample of the United States population. Drug Alcohol Depend. 2017;180:204–207. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Weinberger AH, Pacek LR, Giovenco D, et al. Cigarette use among individuals with alcohol use disorders in the United States, 2002 to 2016: trends overall and by race/ethnicity. Alcohol Clin Exp Res. 2019;43(1):79–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Hughes JR. An update on hardening: a qualitative review. Nicotine Tob Res. 2020;22(6):867–871. [DOI] [PubMed] [Google Scholar]
- 16. Dermody SS, Hendershot CS. A critical review of the effects of nicotine and alcohol coadministration in human laboratory studies. Alcohol Clin Exp Res. 2017;41(3):473–486. [DOI] [PubMed] [Google Scholar]
- 17. Dermody SS, Donny EC. The predicted impact of reducing the nicotine content in cigarettes on alcohol use. Nicotine Tob Res. 2014;16(8):1033–1044. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. McKee SA, Weinberger AH. How can we use our knowledge of alcohol-tobacco interactions to reduce alcohol use? Annu Rev Clin Psychol. 2013;9:649–674. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Verplaetse TL, McKee SA. An overview of alcohol and tobacco/nicotine interactions in the human laboratory. Am J Drug Alcohol Abuse. 2017;43(2):186–196. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Danaei G, Ding EL, Mozaffarian D, et al. The preventable causes of death in the United States: comparative risk assessment of dietary, lifestyle, and metabolic risk factors. PLoS Med. 2009;6(4):e1000058. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Hatsukami DK, Perkins KA, Lesage MG, et al. Nicotine reduction revisited: science and future directions. Tob Control. 2010;19(5):e1–e10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Hatsukami DK, Heishman SJ, Vogel RI, et al. Dose–response effects of spectrum research cigarettes. Nicotine Tob Res. 2013;15(6):1113–1121. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Buchhalter AR, Acosta MC, Evans SE, Breland AB, Eissenberg T. Tobacco abstinence symptom suppression: the role played by the smoking-related stimuli that are delivered by denicotinized cigarettes. Addiction. 2005;100(4):550–559. [DOI] [PubMed] [Google Scholar]
- 24. Tan X, Shiyko MP, Li R, Li Y, Dierker L. A time-varying effect model for intensive longitudinal data. Psychol Methods. 2012;17(1):61–77. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Richter P, Steven PR, Bravo R, et al. Characterization of SPECTRUM Variable Nicotine Research Cigarettes. Tob Regul Sci. 2016;2(2):94–105. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Heatherton TF, Kozlowski LT, Frecker RC, Fagerström KO. The Fagerström Test for Nicotine Dependence: a revision of the Fagerström Tolerance Questionnaire. Br J Addict. 1991;86(9):1119–1127. [DOI] [PubMed] [Google Scholar]
- 27. Selzer ML, Vinokur A, van Rooijen L. A self-administered Short Michigan Alcoholism Screening Test (SMAST). J Stud Alcohol. 1975;36(1):117–126. [DOI] [PubMed] [Google Scholar]
- 28. Sobell LC, Sobell MB. Timeline follow-back. In: Allen J, Litten RZ, eds. Measuring Alcohol Consumption: Psychosocial and Biological Methods. Totowa, NJ: Humana Press; 1992:41–72. [Google Scholar]
- 29. R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2019. [Google Scholar]
- 30. Molenberghs G, Bijnens L, Shaw D. Linear mixed models and missing data. In: Linear Mixed Models in Practice. New York, NY: Springer; 1997:191–274. [Google Scholar]
- 31. Hedden SL, Woolson RF, Carter RE, Palesch Y, Upadhyaya HP, Malcolm RJ. The impact of loss to follow-up on hypothesis tests of the treatment effect for several statistical methods in substance abuse clinical trials. J Subst Abuse Treat. 2009;37(1):54–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Verbeke G. Linear mixed models for longitudinal data. In: Linear Mixed Models in Practice. New York, NY: Springer; 1997:63–153. [Google Scholar]
- 33. Li R, Dziak JD, Tan X, Huang L, Wagner AT, Yang J.. TVEM (Time-Varying Effect Model) SAS Macro Users’ Guide (Version 3.1.1). University Park, PA: The Methodology Center, Penn State; 2017. [Google Scholar]
- 34. Hashibe M, Brennan P, Benhamou S, et al. Alcohol drinking in never users of tobacco, cigarette smoking in never drinkers, and the risk of head and neck cancer: pooled analysis in the International Head and Neck Cancer Epidemiology Consortium. J Natl Cancer Inst. 2007;99(10):777–789. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.