- Split View
-
Views
-
Cite
Cite
Amanda R Mathew, Elizabeth Garrett-Mayer, Bryan W Heckman, Amy E Wahlquist, Matthew J Carpenter, One-Year Smoking Trajectories Among Established Adult Smokers With Low Baseline Motivation to Quit, Nicotine & Tobacco Research, Volume 20, Issue 1, January 2018, Pages 50–57, https://doi.org/10.1093/ntr/ntw264
- Share Icon Share
Abstract
The majority of smokers do not intend to quit in the near term, making unmotivated smokers a key group to target in public health efforts. Although it is often assumed that continuing smokers will have stable rates of smoking over time, limited research has addressed this issue, particularly among smokers not seeking treatment. In the current study, the aims were to (1) characterize the trajectory of naturalistic smoking among unmotivated smokers and (2) examine relationships between naturalistic smoking trajectories and other smoking-related variables.
The study sample comprised 579 control-group (ie, untreated) smokers within a parent clinical trial, who completed a total of nine assessments over 1 year.
Trajectory modeling identified four smoking trajectory groups: stable (72%), shallow decreasers (20%), steep decreasers (5%), and increasers (3%). Membership in the decreasing groups was associated with higher motivation to quit, greater history of quit attempts, and higher cigarettes per day. Females were more likely to be in the increasing versus stable group.
Findings provide needed information on stability and change in cigarette consumption over the course of 1 year among an untreated sample of smokers and identified baseline sociodemographic and smoking-related predictors of smoking trajectory group. Refining understanding of these groups is critical in updating population-based tobacco policy modeling efforts and informing cessation induction efforts that capitalize on naturalistic changes in smoking rate over time.
In the current study, we found that approximately 25% of smokers who endorsed low quit motivation at baseline reduced their cigarette consumption over the course of a year, while 3% increased their cigarette consumption and the majority of smokers (72%) maintained a stable pattern. Refining understanding of smoking trajectories is critical in updating population-based tobacco policy modeling efforts and informing cessation induction efforts that capitalize on naturalistic changes in smoking rate over time.
Introduction
The majority of persistent daily adult smokers do not intend to quit smoking within the next 30 days, and only 50% make a quit attempt in any given year,1 making unmotivated smokers (ie, those not ready to quit in the near term) a key group to target in public health efforts. Although a large literature has addressed smoking trajectories among adolescents,2,3 little empirical research has examined the temporal stability of cigarette consumption among established smokers. On one hand, it is often assumed that smoking among untreated smokers remains largely stable over time, in the absence of a quit attempt. This is consistent with established models of addiction, which posit that the pharmacologic effects of nicotine produce dependence and sustain ongoing smoking.4 In line with this account, smoking is thought to occur at a consistent rate in order to maintain steady nicotine titration.5,6 Negative reinforcement theories of nicotine dependence7 posit that smoking is maintained in order to avoid or escape aversive symptoms of withdrawal caused by periods of abstinence. As nicotine half-life is relatively brief,8 withdrawal symptoms occur frequently among established smokers (ie, within 30 minutes of last cigarette),9 prompting continuous smoking. Thus, these models suggest that among established smokers, especially those unmotivated to quit, smoking rate should remain consistent over time.
On the other hand, motivation to quit or reduce smoking is known to change over time, suggesting that smoking may fluctuate as well. The few studies that have addressed this question have identified a trend toward significant instability in smoking rate. In community samples of smokers motivated to quit, frequent shifts between abstinence, reduced smoking, and heavy smoking over a period of 1–3 months have been documented.10–12 Smoking reduction is a common trend among established smokers, and occurs more often than quit attempts among smokers in the general population.13,14 In a sample of 57 smokers followed for 40 days after the completion of a smoking reduction program, 40% of participants further decreased cigarette consumption while 47% increased, and only 12% maintained a consistent smoking rate.15 Lastly, Yong and colleagues16 identified an overall decline in cigarette consumption among a sample of established cigarette smokers followed for 2–5 years and assessed annually.
Taken together, findings show that fluctuations in smoking rate are much more common than posited by many theoretical models of smoking addiction, highlighting the need for further study of naturalistic changes in smoking behavior and cessation-related variables. Furthermore, the time frame for previous studies of smoking trajectory tends to be either very brief with frequent, fine grain measurements (ie, 28 days, assessed daily)10,11 or very long with infrequent measurement (ie, 2–5 years, assessed annually).16 The duration and frequency of assessment can impact our understanding of smoking consistency. Our study adds to this literature through assessment of smoking behavior among a national sample of established smokers across a follow-up period of 12 months with relatively frequent assessments. Our aims were to (1) describe trajectories of cigarette consumption over the course of 1 year, as well as group-level and participant-level variability associated with these trajectories, and (2) test sociodemographic and smoking-related characteristics as potential predictors of smoking trajectories.
Methods
Overview
Data were drawn from a larger clinical trial in which adults smokers were randomly assigned to either receive a smokeless tobacco product to sample ad libitum for 6 weeks, or receive no intervention.17 All participants were provided with a brief (<3 minutes) prompt to quit, inclusive of referral to state quitline, 6 weeks after study enrollment, followed by 1 year of assessment. The current study is restricted to only those participants assigned to the control condition (N = 610) who did not receive any tobacco products, thus affording an opportunity to examine 1-year patterns of smoking among a large, nationwide sample of smokers. Participants completed a total of nine phone-based assessments over the course of 1 year (weeks 0, 3, 6, 10, 14, 18, 32, 46, and 58). All procedures were approved by appropriate regulatory review.
Participants
Using two national market research panels (Survey Sampling International and Knowledge Networks), potential participants were e-mailed with study invitation. Eligibility was determined through an online link, and those who passed and expressed interest in study participation were mailed a packet with consent form, baseline questionnaire, and return address envelope (N = 2785). Those who returned a study packet/questionnaire and were reached for initial call (N = 1236; 44.4%) were formally enrolled into the trial. All remaining follow-up assessments were conducted by phone.
Eligible participants met the following inclusion criteria: English-speaking, lived in the United States, smoked >10 cigarettes per day (CPD), >19 years old, and unmotivated to quit smoking, as defined by (1) <7 on a 0–10 contemplation ladder (see description below) and (2) no intention to quit in next 30 days. Exclusion criteria were: use of smokeless tobacco on more than one occasion during the past 6 months, breastfeeding/pregnant/planning to become pregnant, any cardiovascular trauma in the past 6 months, quit attempt lasting more than 1 week during the past 6 months, and use of any pharmacotherapy for cessation during the past 3 months.
Measures
Baseline Questionnaire
The baseline questionnaire assessed sociodemographic factors and smoking-related characteristics with items drawn from the Centers for Disease Control and Prevention National Health Interview Survey (NHIS) and National Adult Tobacco Survey (NATS). Nicotine dependence was assessed with the Heaviness of Smoking Index.18
Motivation and Confidence to Quit
Cessation readiness indices were assessed at baseline (ie, week 0) with a modified Contemplation Ladder19,20 to measure (1) interest in quitting and (2) confidence in ability to quit in the next month. Scores ranged from 0 to 10, with higher scores indicating greater motivation/confidence to quit.
Cigarette Consumption
At each assessment, participants were asked if they had used any cigarette (even a puff) or other tobacco product since the previous phone call. For those who reported any use, quantity was assessed via Timeline Follow Back (TLFB) methods.21 For a more robust indicator of smoking, we averaged CPD over the preceding 7 days at each assessment.
Analytic Strategy
To characterize and identify longitudinal smoking trajectories, group-based trajectory modeling was performed.22–24 Briefly, the model assumes that there are a finite set of groups in the population and each group is characterized by a trajectory: in this case, the trajectory is represented by the trend in percent change in CPD over the course of the study. The dependent variable in the model is the percentage change in CPD from baseline (ie, where i represents patient, t represents time, and t = 0 is baseline [week 0]). The approach identifies clusters of individuals using finite mixture modeling, and assigns each individual probabilities of belonging to each group, based on his/her observed trajectory. We have implemented the approach in Stata version 12.1 using the “traj” command.22 A series of models were fit, with the number of clusters ranging from two to four. In each model, we assumed there was an “intercept only” group (ie, a cluster of individuals who had no change in CPD from baseline). Iterative modeling was performed to determine the shape of the trajectories for other clusters, where linear, quadratic, and cubic spline patterns were considered. Bayesian information criteria (BIC) was used to select among the models considered. The best-fitting model included four clusters: (1) stable (intercept only), (2) shallow decreasing (quadratic), (3) steep decreasing (cubic), and (4) shallow increasing (cubic). The resulting model was plotted using R software (R version 3.3.0 [May 2016]; Figure 1).
Secondarily, the proportion of participants who reported a 20% increase and the proportion of participants who reported 50% decrease in CPD relative to baseline at any point were examined. A 50% reduction in CPD was chosen as the threshold for decreased usage, as this is considered clinically significant25 and is the most commonly reported outcome within the smoking reduction literature.26,27 There is no such metric for increased usage, so 20% was chosen as this lower value allows for a greater degree of sensitivity in detecting a reliable smoking increase that may be associated with greater health risk.
Predictors of the smoking trajectory group membership were explored by testing between-group differences in baseline demographic and smoking-related characteristics with one-way ANOVAs for continuous variables and chi-square tests for categorical variables. All baseline variables that showed a significant relationship at the α = 0.20 level for between-group comparisons were evaluated as potential predictors of group membership. A series of logistic regression models were then used to predict group membership with those in the stable group serving as the reference (ie, odds ratios for steep decreasing vs. stable; shallow decreasing vs. stable; and shallow increasing vs. stable are reported). In order to address issues of multicollinearity, baseline variables were examined both as univariate predictors and in a full, multivariable model utilizing backward stepwise elimination. The resulting reduced model, including only covariates that showed a significant relationship at the α = 0.05 level, is reported herein. Comparisons across trajectory groups were performed in R version 3.3.0 (May 2016).
Missing Data
Due to the need for sufficient number of assessments for determining patterns, participants who completed less than five assessments (4.9%) were excluded, leaving a final sample size of N = 579 for all following analyses. Each participant’s trajectory was fit to available data, with 70% of participants completing all nine visits, while 16% completed eight visits, 5% completed seven visits, 6% completed six visits, and 3% completed five visits. No imputation of missing data was performed.
Results
Preliminary Analyses
Overall, the mean age of the sample was 49.3 (SD = 12.4), and participants were 64.3% female. Race/ethnicity was identified as white, non-Hispanic (86.6%), African American (10.5%), or other (2.6%). The majority were unemployed (55.7%), and 44.3% of participants were employed part-time or full time. Mean CPD at baseline was 20.0 (SD = 8.4) with a mean of 32.1 years smoking (median = 35.0; SD = 12.8). The average Heaviness of Smoking Index score was 3.5 (SD = 1.2), indicating moderate nicotine dependence. On average, participants reported 2.8 lifetime quit attempts (median = 2.0; SD = 4.1). Over the course of the study (weeks 0–58), the majority of participants reported no attempt to quit smoking (69.5%, n = 403), while 30.5% of participants (n = 177) made one or more attempts to quit. Of those participants who attempted to quit, 54.2% (n = 96) made one quit attempt, 22.6% (n = 40) made two quit attempts, and 23.2% (n = 41) made three or more quit attempts. Following the brief prompt to quit (ie, from week 10 through week 58), 3.6% of participants (n = 21) used a quitline or phone-based counseling service, and 25.0% (n = 145) used other smoking cessation resources (eg, counseling, pharmacotherapy, or use of internet-based resources).
Smoking Trajectory Group
Group Assignment
Trajectory modeling identified four groups based on the fitted trajectory of participants’ CPD over time. As shown in Figure 1, the most prevalent pattern was stable (n = 416; 72%), followed by shallow decreasers (n = 117; 20%), steep decreasers (n = 27; 5%), and increasers (n = 19; 3%). Those in the steep decreasing group were most likely to make a quit attempt over the course of the study (81.5%), as compared to those in the shallow decreasing (61.5%), stable (19.2%), and increasing groups (15.8%).
Secondarily, the number of assessments in which participants reported decreased CPD (≤50%) or increased CPD (≥20%) relative to baseline were examined. Overall, the majority of participants (51.0%) did not report any increase or decrease in CPD as assessed with these thresholds, while 5.5% of participants reported both a significant increase from baseline and a significant decrease in CPD from baseline at different time points over the course of the study. Approximately 18.4% of participants reported only an increase, while 25.0% reported only a decrease in CPD. Within smoking trajectory groups reported above, 93% of steep decreasers reported a clinically-significant smoking reduction of this threshold over the course of the study, while 68% of shallow decreasers reported a decrease meeting this threshold. Within the stable group, 65% of participants reported neither an increase nor decrease associated with the thresholds above, while 21% reported only an increase, 10% reported only a decrease, and 5% reported both. All of those in the increasing group endorsed an increase of CPD ≥ 20%.
Predictors of Smoking Rate Trajectories
Baseline demographic and smoking-related characteristics by smoking group are presented in Table 1. Significant between-group differences were identified for six baseline variables: gender, CPD, nicotine dependence, prior quit attempts, health concerns, and motivation to quit. Pairwise comparisons indicated that CPD and motivation to quit values were higher among those in both decreasing groups versus stable group. Those in the steep decreasing group had greater levels of nicotine dependence and more prior quit attempts (≥6), as compared to the stable group. There was a higher proportion of females in the increasing versus stable group. Lastly, the shallow decreasing group had a higher proportion of those endorsing no health concerns versus the stable group.
. | Smoking group . | . | |||
---|---|---|---|---|---|
. | Steep decreasing (n = 27) . | Shallow decreasing (n = 117) . | Stable (n = 416) . | Shallow increasing (n = 19) . | p . |
Female | 44% | 63% | 65% | 89%a | .016 |
Race/ethnicity | .53 | ||||
Non-Hispanic Caucasian | 93% | 85% | 87% | 78% | |
African American | 7% | 14% | 10% | 17% | |
Other | 0% | 2% | 3% | 6% | |
Age in years | 47.0 (12.9) | 50.5 (13.9) | 49.0 (11.9) | 50.7 (13.3) | .54 |
Married or member of couple | 67% | 49% | 51% | 32% | .13 |
Education | .99 | ||||
<HS diploma | 41% | 38% | 38% | 37% | |
Some college/ technical school | 41% | 46% | 43% | 47% | |
> College degree | 19% | 16% | 19% | 16% | |
Annual household income | .43 | ||||
<$25 000 | 26% | 35% | 36% | 47% | |
$25 000–$50 000 | 48% | 30% | 34% | 37% | |
>$50 000 | 26% | 35% | 30% | 16% | |
Employed full- or part-time | 48% | 38% | 46% | 32% | .26 |
Years smoking | 30.5 (12.3) | 33.3 (14.5) | 31.9 (12.2) | 33.2 (14.2) | .69 |
Cigarettes per day | 28.9 (9.0)b | 21.2 (8.3)b | 19.0 (8.0) | 22.0 (7.9) | <.001 |
Nicotine dependence | 4.2 (1.2)b | 3.5 (1.2) | 3.4 (1.2) | 3.7 (1.3) | .018 |
Prior quit attempts | <.01 | ||||
0 | 7% | 18% | 20% | 16% | |
1 | 15% | 20% | 17% | 21% | |
2–5 | 48% | 49% | 56% | 58% | |
>6 | 30%a | 14% | 7% | 5% | |
Made quit attempt in past 12 mo | 15% | 12% | 7% | 11% | .11 |
Health concerns | .032 | ||||
Somewhat/very concerned | 85% | 68% | 75% | 68% | |
Slightly concerned | 11% | 18% | 21% | 26% | |
Not concerned | 4% | 14%b | 5% | 5% | |
Provider advised to quit | 85% | 73% | 76% | 79 | .62 |
Live with another smoker | 59% | 44% | 47% | 42 | .56 |
Smoking-related health condition | 37% | 33% | 33% | 42% | .79 |
Motivation to quit (0–10) | 2.41 (3.15)b | 1.67 (2.55)b | 1.14 (2.22) | 1.58 (2.67) | .013 |
Confidence to quit (0–10) | 2.22 (2.97) | 3.09 (3.14) | 2.67 (2.99) | 2.74 (2.79) | .45 |
. | Smoking group . | . | |||
---|---|---|---|---|---|
. | Steep decreasing (n = 27) . | Shallow decreasing (n = 117) . | Stable (n = 416) . | Shallow increasing (n = 19) . | p . |
Female | 44% | 63% | 65% | 89%a | .016 |
Race/ethnicity | .53 | ||||
Non-Hispanic Caucasian | 93% | 85% | 87% | 78% | |
African American | 7% | 14% | 10% | 17% | |
Other | 0% | 2% | 3% | 6% | |
Age in years | 47.0 (12.9) | 50.5 (13.9) | 49.0 (11.9) | 50.7 (13.3) | .54 |
Married or member of couple | 67% | 49% | 51% | 32% | .13 |
Education | .99 | ||||
<HS diploma | 41% | 38% | 38% | 37% | |
Some college/ technical school | 41% | 46% | 43% | 47% | |
> College degree | 19% | 16% | 19% | 16% | |
Annual household income | .43 | ||||
<$25 000 | 26% | 35% | 36% | 47% | |
$25 000–$50 000 | 48% | 30% | 34% | 37% | |
>$50 000 | 26% | 35% | 30% | 16% | |
Employed full- or part-time | 48% | 38% | 46% | 32% | .26 |
Years smoking | 30.5 (12.3) | 33.3 (14.5) | 31.9 (12.2) | 33.2 (14.2) | .69 |
Cigarettes per day | 28.9 (9.0)b | 21.2 (8.3)b | 19.0 (8.0) | 22.0 (7.9) | <.001 |
Nicotine dependence | 4.2 (1.2)b | 3.5 (1.2) | 3.4 (1.2) | 3.7 (1.3) | .018 |
Prior quit attempts | <.01 | ||||
0 | 7% | 18% | 20% | 16% | |
1 | 15% | 20% | 17% | 21% | |
2–5 | 48% | 49% | 56% | 58% | |
>6 | 30%a | 14% | 7% | 5% | |
Made quit attempt in past 12 mo | 15% | 12% | 7% | 11% | .11 |
Health concerns | .032 | ||||
Somewhat/very concerned | 85% | 68% | 75% | 68% | |
Slightly concerned | 11% | 18% | 21% | 26% | |
Not concerned | 4% | 14%b | 5% | 5% | |
Provider advised to quit | 85% | 73% | 76% | 79 | .62 |
Live with another smoker | 59% | 44% | 47% | 42 | .56 |
Smoking-related health condition | 37% | 33% | 33% | 42% | .79 |
Motivation to quit (0–10) | 2.41 (3.15)b | 1.67 (2.55)b | 1.14 (2.22) | 1.58 (2.67) | .013 |
Confidence to quit (0–10) | 2.22 (2.97) | 3.09 (3.14) | 2.67 (2.99) | 2.74 (2.79) | .45 |
For categorical variables, percentages for each response are provided and p values were determined by Pearson’s chi-square tests. For continuous variables, mean and standard deviation values are provided and p values were determined by one-way ANOVA tests. Motivation to quit and confidence to quit were assessed at the initial study phone call (ie, week 0). p values < .05 are indicated in bold font.
aIndicates statistically different values from the stable group from pairwise comparisons of column proportions using a Bonferroni correction in Pearson’s chi-square tests.
bIndicates statistically different values from the stable group from least squares differences for pairwise comparisons in one-way ANOVA tests.
. | Smoking group . | . | |||
---|---|---|---|---|---|
. | Steep decreasing (n = 27) . | Shallow decreasing (n = 117) . | Stable (n = 416) . | Shallow increasing (n = 19) . | p . |
Female | 44% | 63% | 65% | 89%a | .016 |
Race/ethnicity | .53 | ||||
Non-Hispanic Caucasian | 93% | 85% | 87% | 78% | |
African American | 7% | 14% | 10% | 17% | |
Other | 0% | 2% | 3% | 6% | |
Age in years | 47.0 (12.9) | 50.5 (13.9) | 49.0 (11.9) | 50.7 (13.3) | .54 |
Married or member of couple | 67% | 49% | 51% | 32% | .13 |
Education | .99 | ||||
<HS diploma | 41% | 38% | 38% | 37% | |
Some college/ technical school | 41% | 46% | 43% | 47% | |
> College degree | 19% | 16% | 19% | 16% | |
Annual household income | .43 | ||||
<$25 000 | 26% | 35% | 36% | 47% | |
$25 000–$50 000 | 48% | 30% | 34% | 37% | |
>$50 000 | 26% | 35% | 30% | 16% | |
Employed full- or part-time | 48% | 38% | 46% | 32% | .26 |
Years smoking | 30.5 (12.3) | 33.3 (14.5) | 31.9 (12.2) | 33.2 (14.2) | .69 |
Cigarettes per day | 28.9 (9.0)b | 21.2 (8.3)b | 19.0 (8.0) | 22.0 (7.9) | <.001 |
Nicotine dependence | 4.2 (1.2)b | 3.5 (1.2) | 3.4 (1.2) | 3.7 (1.3) | .018 |
Prior quit attempts | <.01 | ||||
0 | 7% | 18% | 20% | 16% | |
1 | 15% | 20% | 17% | 21% | |
2–5 | 48% | 49% | 56% | 58% | |
>6 | 30%a | 14% | 7% | 5% | |
Made quit attempt in past 12 mo | 15% | 12% | 7% | 11% | .11 |
Health concerns | .032 | ||||
Somewhat/very concerned | 85% | 68% | 75% | 68% | |
Slightly concerned | 11% | 18% | 21% | 26% | |
Not concerned | 4% | 14%b | 5% | 5% | |
Provider advised to quit | 85% | 73% | 76% | 79 | .62 |
Live with another smoker | 59% | 44% | 47% | 42 | .56 |
Smoking-related health condition | 37% | 33% | 33% | 42% | .79 |
Motivation to quit (0–10) | 2.41 (3.15)b | 1.67 (2.55)b | 1.14 (2.22) | 1.58 (2.67) | .013 |
Confidence to quit (0–10) | 2.22 (2.97) | 3.09 (3.14) | 2.67 (2.99) | 2.74 (2.79) | .45 |
. | Smoking group . | . | |||
---|---|---|---|---|---|
. | Steep decreasing (n = 27) . | Shallow decreasing (n = 117) . | Stable (n = 416) . | Shallow increasing (n = 19) . | p . |
Female | 44% | 63% | 65% | 89%a | .016 |
Race/ethnicity | .53 | ||||
Non-Hispanic Caucasian | 93% | 85% | 87% | 78% | |
African American | 7% | 14% | 10% | 17% | |
Other | 0% | 2% | 3% | 6% | |
Age in years | 47.0 (12.9) | 50.5 (13.9) | 49.0 (11.9) | 50.7 (13.3) | .54 |
Married or member of couple | 67% | 49% | 51% | 32% | .13 |
Education | .99 | ||||
<HS diploma | 41% | 38% | 38% | 37% | |
Some college/ technical school | 41% | 46% | 43% | 47% | |
> College degree | 19% | 16% | 19% | 16% | |
Annual household income | .43 | ||||
<$25 000 | 26% | 35% | 36% | 47% | |
$25 000–$50 000 | 48% | 30% | 34% | 37% | |
>$50 000 | 26% | 35% | 30% | 16% | |
Employed full- or part-time | 48% | 38% | 46% | 32% | .26 |
Years smoking | 30.5 (12.3) | 33.3 (14.5) | 31.9 (12.2) | 33.2 (14.2) | .69 |
Cigarettes per day | 28.9 (9.0)b | 21.2 (8.3)b | 19.0 (8.0) | 22.0 (7.9) | <.001 |
Nicotine dependence | 4.2 (1.2)b | 3.5 (1.2) | 3.4 (1.2) | 3.7 (1.3) | .018 |
Prior quit attempts | <.01 | ||||
0 | 7% | 18% | 20% | 16% | |
1 | 15% | 20% | 17% | 21% | |
2–5 | 48% | 49% | 56% | 58% | |
>6 | 30%a | 14% | 7% | 5% | |
Made quit attempt in past 12 mo | 15% | 12% | 7% | 11% | .11 |
Health concerns | .032 | ||||
Somewhat/very concerned | 85% | 68% | 75% | 68% | |
Slightly concerned | 11% | 18% | 21% | 26% | |
Not concerned | 4% | 14%b | 5% | 5% | |
Provider advised to quit | 85% | 73% | 76% | 79 | .62 |
Live with another smoker | 59% | 44% | 47% | 42 | .56 |
Smoking-related health condition | 37% | 33% | 33% | 42% | .79 |
Motivation to quit (0–10) | 2.41 (3.15)b | 1.67 (2.55)b | 1.14 (2.22) | 1.58 (2.67) | .013 |
Confidence to quit (0–10) | 2.22 (2.97) | 3.09 (3.14) | 2.67 (2.99) | 2.74 (2.79) | .45 |
For categorical variables, percentages for each response are provided and p values were determined by Pearson’s chi-square tests. For continuous variables, mean and standard deviation values are provided and p values were determined by one-way ANOVA tests. Motivation to quit and confidence to quit were assessed at the initial study phone call (ie, week 0). p values < .05 are indicated in bold font.
aIndicates statistically different values from the stable group from pairwise comparisons of column proportions using a Bonferroni correction in Pearson’s chi-square tests.
bIndicates statistically different values from the stable group from least squares differences for pairwise comparisons in one-way ANOVA tests.
For those baseline predictors included in univariate regression models (ie, those with p < .20 in ANOVA or chi-square tests), several significant predictors of group membership were identified, as shown in Table 2.
. | Steep decreasing (n = 27) vs. stable (n = 416) . | Shallow decreasing (n = 117) vs. stable (n = 416) . | Shallow increasing (n = 19) vs. stable (n = 416) . | |||
---|---|---|---|---|---|---|
. | OR (95% CI) . | OR (95% CI) . | OR (95% CI) . | |||
. | Univariate . | Reduced . | Univariate . | Reduced . | Univariate . | Reduced . |
Female | 0.44 (0.20, 0.97)* | 0.47 (0.19, 1.14)** | 0.94 (0.61, 1.45) | 4.65 (1.03, 21.0)* | 5.32 (1.20, 23.7)* | |
Racea | ||||||
Non-Hispanic Caucasian | ref | ref | ref | |||
African American | 0.90 (0.54, 1.49) | 1.14 (0.92, 1.40) | 1.25 (0.81, 1.94) | |||
Income | ||||||
>$50 000 | ref | ref | ref | |||
$25 000–$50 000 | 1.60 (0.61, 4.2) | 0.73 (0.44, 1.24) | 2.01 (0.49, 8.15) | |||
<$25 000 | 0.82 (0.27, 2.45) | 0.82 (0.50, 1.46) | 2.46 (0.63, 9.54) | |||
Employed | 1.07 (0.48, 2.38) | 0.72 (0.47, 1.11) | 0.53 (0.20, 1.46) | |||
Cigarettes per day (5 units) | 1.76 (1.42, 2.18)* | 1.20 (1.10, 1.30)* | 1.17 (1.03, 1.32) | 1.07 (1.02, 1.11)* | 1.23 (0.95, 1.59) | 1.05 (0.99, 1.11)** |
HSI score (1 unit) | 1.70 (1.19, 2.43)* | 0.58 (0.31, 1.08)** | 1.08 (0.91, 1.28) | 0.79 (0.60, 1.03)** | 1.17 (0.79, 1.74) | |
Prior quit attempts | ||||||
0 | ref | ref | ref | ref | ref | |
1 | 2.37 (0.41, 13.8) | 1.32 (0.18, 9.46) | 1.30 (0.65, 2.57) | 1.36 (0.67, 2.75) | 1.58 (0.33, 7.52) | |
2–5 | 2.34 (0.50, 10.93) | 2.39 (0.49, 11.61) | 0.98 (0.55, 1.73) | 1.11 (0.61, 2.03) | 1.32 (0.35, 4.99) | |
>6 | 12.0 (2.33, 61.88)* | 9.06 (1.59, 51.5)* | 2.29 (1.03, 5.06)* | 2.35 (1.03, 5.37)* | 1.00 (0.10, 10.49) | |
Health concerns | ||||||
Somewhat/very | ref | ref | ref | ref | ||
Slightly | 0.47 (0.13, 1.63) | 0.94 (0.54, 1.62) | 1.06 (0.61, 1.86) | 1.37 (0.47, 4.04) | ||
Not concerned | 0.71 (0.09, 5.78) | 3.26 (1.58, 6.73)* | 3.77 (1.77, 8.05)* | 1.26 (0.15, 10.55) | ||
Motivation to quit (2 units) | 1.41 (1.09, 1.83)* | 1.21 (1.03, 1.42)* | 1.20 (1.01, 1.41)* | 1.12 (1.03, 1.22)* | 1.16 (0.81, 1.66) |
. | Steep decreasing (n = 27) vs. stable (n = 416) . | Shallow decreasing (n = 117) vs. stable (n = 416) . | Shallow increasing (n = 19) vs. stable (n = 416) . | |||
---|---|---|---|---|---|---|
. | OR (95% CI) . | OR (95% CI) . | OR (95% CI) . | |||
. | Univariate . | Reduced . | Univariate . | Reduced . | Univariate . | Reduced . |
Female | 0.44 (0.20, 0.97)* | 0.47 (0.19, 1.14)** | 0.94 (0.61, 1.45) | 4.65 (1.03, 21.0)* | 5.32 (1.20, 23.7)* | |
Racea | ||||||
Non-Hispanic Caucasian | ref | ref | ref | |||
African American | 0.90 (0.54, 1.49) | 1.14 (0.92, 1.40) | 1.25 (0.81, 1.94) | |||
Income | ||||||
>$50 000 | ref | ref | ref | |||
$25 000–$50 000 | 1.60 (0.61, 4.2) | 0.73 (0.44, 1.24) | 2.01 (0.49, 8.15) | |||
<$25 000 | 0.82 (0.27, 2.45) | 0.82 (0.50, 1.46) | 2.46 (0.63, 9.54) | |||
Employed | 1.07 (0.48, 2.38) | 0.72 (0.47, 1.11) | 0.53 (0.20, 1.46) | |||
Cigarettes per day (5 units) | 1.76 (1.42, 2.18)* | 1.20 (1.10, 1.30)* | 1.17 (1.03, 1.32) | 1.07 (1.02, 1.11)* | 1.23 (0.95, 1.59) | 1.05 (0.99, 1.11)** |
HSI score (1 unit) | 1.70 (1.19, 2.43)* | 0.58 (0.31, 1.08)** | 1.08 (0.91, 1.28) | 0.79 (0.60, 1.03)** | 1.17 (0.79, 1.74) | |
Prior quit attempts | ||||||
0 | ref | ref | ref | ref | ref | |
1 | 2.37 (0.41, 13.8) | 1.32 (0.18, 9.46) | 1.30 (0.65, 2.57) | 1.36 (0.67, 2.75) | 1.58 (0.33, 7.52) | |
2–5 | 2.34 (0.50, 10.93) | 2.39 (0.49, 11.61) | 0.98 (0.55, 1.73) | 1.11 (0.61, 2.03) | 1.32 (0.35, 4.99) | |
>6 | 12.0 (2.33, 61.88)* | 9.06 (1.59, 51.5)* | 2.29 (1.03, 5.06)* | 2.35 (1.03, 5.37)* | 1.00 (0.10, 10.49) | |
Health concerns | ||||||
Somewhat/very | ref | ref | ref | ref | ||
Slightly | 0.47 (0.13, 1.63) | 0.94 (0.54, 1.62) | 1.06 (0.61, 1.86) | 1.37 (0.47, 4.04) | ||
Not concerned | 0.71 (0.09, 5.78) | 3.26 (1.58, 6.73)* | 3.77 (1.77, 8.05)* | 1.26 (0.15, 10.55) | ||
Motivation to quit (2 units) | 1.41 (1.09, 1.83)* | 1.21 (1.03, 1.42)* | 1.20 (1.01, 1.41)* | 1.12 (1.03, 1.22)* | 1.16 (0.81, 1.66) |
CI = confidence interval; HSI = Heaviness of Smoking Index; OR = odds ratio; ref = reference group.
aRace = other removed due to sparseness.
*p< .05; **p < .10.
. | Steep decreasing (n = 27) vs. stable (n = 416) . | Shallow decreasing (n = 117) vs. stable (n = 416) . | Shallow increasing (n = 19) vs. stable (n = 416) . | |||
---|---|---|---|---|---|---|
. | OR (95% CI) . | OR (95% CI) . | OR (95% CI) . | |||
. | Univariate . | Reduced . | Univariate . | Reduced . | Univariate . | Reduced . |
Female | 0.44 (0.20, 0.97)* | 0.47 (0.19, 1.14)** | 0.94 (0.61, 1.45) | 4.65 (1.03, 21.0)* | 5.32 (1.20, 23.7)* | |
Racea | ||||||
Non-Hispanic Caucasian | ref | ref | ref | |||
African American | 0.90 (0.54, 1.49) | 1.14 (0.92, 1.40) | 1.25 (0.81, 1.94) | |||
Income | ||||||
>$50 000 | ref | ref | ref | |||
$25 000–$50 000 | 1.60 (0.61, 4.2) | 0.73 (0.44, 1.24) | 2.01 (0.49, 8.15) | |||
<$25 000 | 0.82 (0.27, 2.45) | 0.82 (0.50, 1.46) | 2.46 (0.63, 9.54) | |||
Employed | 1.07 (0.48, 2.38) | 0.72 (0.47, 1.11) | 0.53 (0.20, 1.46) | |||
Cigarettes per day (5 units) | 1.76 (1.42, 2.18)* | 1.20 (1.10, 1.30)* | 1.17 (1.03, 1.32) | 1.07 (1.02, 1.11)* | 1.23 (0.95, 1.59) | 1.05 (0.99, 1.11)** |
HSI score (1 unit) | 1.70 (1.19, 2.43)* | 0.58 (0.31, 1.08)** | 1.08 (0.91, 1.28) | 0.79 (0.60, 1.03)** | 1.17 (0.79, 1.74) | |
Prior quit attempts | ||||||
0 | ref | ref | ref | ref | ref | |
1 | 2.37 (0.41, 13.8) | 1.32 (0.18, 9.46) | 1.30 (0.65, 2.57) | 1.36 (0.67, 2.75) | 1.58 (0.33, 7.52) | |
2–5 | 2.34 (0.50, 10.93) | 2.39 (0.49, 11.61) | 0.98 (0.55, 1.73) | 1.11 (0.61, 2.03) | 1.32 (0.35, 4.99) | |
>6 | 12.0 (2.33, 61.88)* | 9.06 (1.59, 51.5)* | 2.29 (1.03, 5.06)* | 2.35 (1.03, 5.37)* | 1.00 (0.10, 10.49) | |
Health concerns | ||||||
Somewhat/very | ref | ref | ref | ref | ||
Slightly | 0.47 (0.13, 1.63) | 0.94 (0.54, 1.62) | 1.06 (0.61, 1.86) | 1.37 (0.47, 4.04) | ||
Not concerned | 0.71 (0.09, 5.78) | 3.26 (1.58, 6.73)* | 3.77 (1.77, 8.05)* | 1.26 (0.15, 10.55) | ||
Motivation to quit (2 units) | 1.41 (1.09, 1.83)* | 1.21 (1.03, 1.42)* | 1.20 (1.01, 1.41)* | 1.12 (1.03, 1.22)* | 1.16 (0.81, 1.66) |
. | Steep decreasing (n = 27) vs. stable (n = 416) . | Shallow decreasing (n = 117) vs. stable (n = 416) . | Shallow increasing (n = 19) vs. stable (n = 416) . | |||
---|---|---|---|---|---|---|
. | OR (95% CI) . | OR (95% CI) . | OR (95% CI) . | |||
. | Univariate . | Reduced . | Univariate . | Reduced . | Univariate . | Reduced . |
Female | 0.44 (0.20, 0.97)* | 0.47 (0.19, 1.14)** | 0.94 (0.61, 1.45) | 4.65 (1.03, 21.0)* | 5.32 (1.20, 23.7)* | |
Racea | ||||||
Non-Hispanic Caucasian | ref | ref | ref | |||
African American | 0.90 (0.54, 1.49) | 1.14 (0.92, 1.40) | 1.25 (0.81, 1.94) | |||
Income | ||||||
>$50 000 | ref | ref | ref | |||
$25 000–$50 000 | 1.60 (0.61, 4.2) | 0.73 (0.44, 1.24) | 2.01 (0.49, 8.15) | |||
<$25 000 | 0.82 (0.27, 2.45) | 0.82 (0.50, 1.46) | 2.46 (0.63, 9.54) | |||
Employed | 1.07 (0.48, 2.38) | 0.72 (0.47, 1.11) | 0.53 (0.20, 1.46) | |||
Cigarettes per day (5 units) | 1.76 (1.42, 2.18)* | 1.20 (1.10, 1.30)* | 1.17 (1.03, 1.32) | 1.07 (1.02, 1.11)* | 1.23 (0.95, 1.59) | 1.05 (0.99, 1.11)** |
HSI score (1 unit) | 1.70 (1.19, 2.43)* | 0.58 (0.31, 1.08)** | 1.08 (0.91, 1.28) | 0.79 (0.60, 1.03)** | 1.17 (0.79, 1.74) | |
Prior quit attempts | ||||||
0 | ref | ref | ref | ref | ref | |
1 | 2.37 (0.41, 13.8) | 1.32 (0.18, 9.46) | 1.30 (0.65, 2.57) | 1.36 (0.67, 2.75) | 1.58 (0.33, 7.52) | |
2–5 | 2.34 (0.50, 10.93) | 2.39 (0.49, 11.61) | 0.98 (0.55, 1.73) | 1.11 (0.61, 2.03) | 1.32 (0.35, 4.99) | |
>6 | 12.0 (2.33, 61.88)* | 9.06 (1.59, 51.5)* | 2.29 (1.03, 5.06)* | 2.35 (1.03, 5.37)* | 1.00 (0.10, 10.49) | |
Health concerns | ||||||
Somewhat/very | ref | ref | ref | ref | ||
Slightly | 0.47 (0.13, 1.63) | 0.94 (0.54, 1.62) | 1.06 (0.61, 1.86) | 1.37 (0.47, 4.04) | ||
Not concerned | 0.71 (0.09, 5.78) | 3.26 (1.58, 6.73)* | 3.77 (1.77, 8.05)* | 1.26 (0.15, 10.55) | ||
Motivation to quit (2 units) | 1.41 (1.09, 1.83)* | 1.21 (1.03, 1.42)* | 1.20 (1.01, 1.41)* | 1.12 (1.03, 1.22)* | 1.16 (0.81, 1.66) |
CI = confidence interval; HSI = Heaviness of Smoking Index; OR = odds ratio; ref = reference group.
aRace = other removed due to sparseness.
*p< .05; **p < .10.
Increasing Versus Stable
Gender was the only baseline predictor associated with increasing group membership, with females more likely to be in the increasing versus stable group.
Shallow Decreasing Versus Stable
Those with six or more lifetime quit attempts were nearly 2.5 times more likely to be in the shallow decreasing group, relative to those with no history of quit attempts. Higher CPD and higher motivation to quit were also associated with membership in the shallow decreasing group.
Steep Decreasing Versus Stable
Females were less likely to be in the steep decreasing versus stable group. Number of prior quit attempts, CPD, nicotine dependence, and motivation to quit were also each positively associated with steep decreasing group membership.
Discussion
Unmotivated smokers are a key target for public health interventions, yet we know little about their smoking maintenance over time. In the current study, we found that approximately 25% of smokers who endorsed low quit motivation at baseline reduced their cigarette consumption over the course of a year, while 3% increased their cigarette consumption and the majority of smokers (72%) maintained a stable pattern. Several baseline predictors were associated with smoking group membership. Greater number of prior quit attempts, greater smoking rate, and higher motivation to quit were associated with decreasing group membership, while female gender was associated with increasing group membership.
Our study identified lower variability in cigarette consumption over the course of 1 year than findings reported in both a study with fine-grained measurement over a shorter assessment period15 and a study with infrequent measurement over a longer period.18 However, we found that approximately one in four smokers decreased CPD from baseline to 12-month follow-up, and 35% of those who were classified as stable overall endorsed short-term fluctuations in CPD. Thus, while our findings are largely consistent with a pharmacological model of nicotine dependence,4 individual differences in smoker characteristics (ie, motivation to quit) and measurement (ie, duration and frequency) may account for differences in longitudinal patterns of smoking behavior observed across studies and play a key role in impacting our understanding of smoking consistency.
As would be expected, smoking reduction was associated with higher motivation to quit and greater history of quit attempts. Together with other studies of longitudinal smoking patterns,11,12,15 our findings underscore the importance of these variables in maintaining reduced smoking rates over time. In the current study, women were less likely to be in the steep decreasing group and overrepresented in the increasing group. This may reflect the presence of gender-specific maintaining factors that are shown to underlie difficulty quitting among women28–30; namely, smoking for greater affect regulation28,31 or weight control motives.32,33 Further study to characterize the role of these factors in relation to naturalistic smoking behavior is needed.
Findings of the current study have important implications for refining our understanding of smoking behavior in smokers not seeking treatment. Although all smokers endorsed low motivation to quit at baseline, roughly 30% of smokers made an attempt to quit over the course of the year (of which, 55% quit for at least 1 week), and incidence of a quit attempt was more than twice as likely for those in the steep or shallow decreasing groups versus stable or increasing groups. Further, one in four of these attempts was treatment-assisted. Undergoing one or more attempts to quit smoking may account, in part, for decreasing trajectory group membership, as those who resume smoking following a quit attempt are more likely to return to a lower level of cigarette consumption.34,35 In the current study, approximately 30% of smokers reported a clinically-significant smoking reduction (ie, ≥50% decrease from baseline value) at one or more timepoints.
Understanding naturalistic smoking patterns among untreated smokers could help guide cessation induction strategies. First, even a small reduction in cigarette smoking could be seen as a positive health behavior change or effort to engage in harm reduction. Reduction in individual health risk from reduced smoking is limited and difficult to demonstrate,36 perhaps due to compensatory smoking behavior among those who smoke fewer CPD. However, reduced smoking is associated with higher rates of downstream quitting.37 Those who have reduced their smoking rate should be encouraged to build on this initial behavior with other strategies to move them closer toward cessation, such as switching to cigarettes with lower nicotine content38 or making a practice quit attempt,39 possibly inclusive of sampling nicotine replacement products (ie, nicotine patch or gum). Second, although an increase in cigarette consumption is typically associated with early stage smokers who are still on the uptake trajectory,2,3 we identified this pattern in a small group of established, adult smokers, which was nearly 90% female. Nearly one-quarter of the sample endorsed smoking increase through a more sensitive indicator using a cut-off score of ≥20% increase from baseline CPD. Further classifying the characteristics and treatment needs of these individuals, and considering gender-specific treatment targets, may be important in targeting interventions for those at risk of worsening tobacco-related health hazards. As women may be less responsive to some commonly-used forms of smoking cessation treatment (ie, nicotine replacement therapy40,41), further innovations in treatment development for fostering smoking cessation among women are needed.42
Several methodological issues warrant comment. First, stability of smoking may be a function of how long and how frequently smoking behavior is assessed, such that increasing duration and frequency of assessments increases the likelihood of finding instability. One strength of the current study is the relative balance of frequency and duration of assessments, as compared to previous studies, allowing us to identify robust trajectories of instability (ie, pattern is maintained over a 12-month time period) that are likely to be of more clinical significance than just day-to-day variation. Second, we cannot rule out the potential impact of reactivity to study assessments on reported smoking rates. Given our observational study design with a minimally-engaged study population, we expect minimal false reports of smoking behavior.43 While it is possible that self-monitoring of smoking rate may alter smoking behavior, a recent study with far more frequent assessments (ie, six assessments daily for 4 weeks) found that frequency of assessment had small effects on self-report of craving and mood, but was unrelated to self-reported smoking outcomes.44,45 Further, we are unable to ascertain if the observed decline in consumption reflects a meaningful change in nicotine dependence/ health risk without biomarker validation, as smokers who reduce CPD could still maintain nicotine intake/carbon monoxide due to compensatory smoking behavior.46 Third, due to smaller group size in the increasing group, we may have been underpowered to detect significant baseline predictors of this trajectory. Lastly, as all participants were provided with a referral to quitline and brief prompt to quit, the current study design is a step away from truly naturalistic. However, we expect this is similar to quit advice smokers commonly receive from healthcare providers47 and rate of quitline uptake in the current study were similar to rates observed in population-based studies (1%–7%).48,49 Further, rates of quit attempts over the course of the study remained relatively low, and findings were similar when those with a period of smoking abstinence were removed from analyses.
In conclusion, there is inconsistency in the literature regarding stability of cigarette consumption over time among established smokers, particularly those with low motivation to quit. In a nationwide sample of smokers not seeking treatment, our study found that 72% of participants maintained a stable smoking pattern over the course of a year, while 25% reduced smoking and 3% increased smoking. We further identified baseline sociodemographic and smoking-related predictors of smoking trajectory group. Refining understanding of these groups is critical in updating population-based tobacco policy modeling efforts and informing cessation induction efforts that capitalize on naturalistic changes in smoking rate over time.
Funding
This work was supported by the National Institutes of Health (R01 CA154992 to MJC, F32 DA036947 to ARM, K12 DA031794 to BWH). EG-M and AEW are supported through the Biostatistics Shared Resource, Hollings Cancer Center, Medical University of South Carolina (P30 CA138313).
Declaration of Interests
None declared.
Acknowledgments
The authors thank Amy Boatright for coordinating data collection.
References
Comments