|Home | About | Journals | Submit | Contact Us | Français|
While declines in smoking prevalence in the United States have been well documented, trends in smokeless tobacco (SLT) use are less clear. This study updates previous analyses of US SLT use prevalence to better understand trends and factors related to SLT use.
We used the Tobacco Use Supplement of the Current Population Survey (TUS-CPS) to examine trends and factors related to SLT use using joinpoint and logistic regression models. SLT consumption from 1985 to 2011 was obtained from the 2011 Federal Trade Commission Smokeless Tobacco Report. Sensitivity analyses were conducted for assessing the impact of varying frequency definitions of SLT use.
Decreasing trends in smoking and SLT prevalence overall were observed from 1992 to 2003 independently of use definition. SLT prevalence in the total adult population significantly decreased at an annual percent change (APC) of 4.5% per year from 1992 to 2003, but has been approximately constant ever since. Similar patterns were also found in adult males (APC = −4.4%) and young males (APC = −9.5%). SLT per capita consumption decreased significantly from 1991 to 1999 (APC = −2.2%), but has since decreased at only 0.35% per year (1999–2011). SLT use was found to be associated with former smoker status, younger age, white race, living in rural areas, residence in the South, lower education and unemployment, adjusting for other factors.
Declines in SLT use were found in the United States, suggesting tobacco control has had positive impacts, but these have slowed since 2003. Targeting tobacco control policies to at-risk demographic groups is needed to further reduce SLT use in the United States.
This study confirms that the declines in SLT use prevalence stopped in 2003 across different demographic groups, consistent with trends in SLT consumption. In addition, the longer period of analysis in comparison with earlier studies allows for quantitative characterization of SLT use trends using joinpoint regression. The study also shows the impact of different SLT use definitions in determining tobacco product use prevalence and trends.
As US smoking prevalence rates have declined, the promotion and consumption of alternative tobacco products, such as smokeless tobacco (SLT) and e-cigarettes, has been increasing.1,2 While the use of SLT was shown to be declining prior to the year 2002,3,4 total SLT sales have been increasing in recent years,2,5,6 likely influenced by the increase in the free nicotine content of SLT7 and the introduction of new forms and delivery systems.8 In particular, use of moist snuff, including snus, currently the largest share of the US market (85.6% in 2011),2 has been increasing since the 1980s. The use of both SLT and cigarettes has been observed in states with weak and strong tobacco control policies. For example, over 10% of smokers also use SLT in states such as Minnesota and Wyoming,9 with substantial recent increases in dual-use in Minnesota between 2007 and 201010 despite strong anti-tobacco policies. This high level of use could represent a substitution of SLT use for cigarettes in places where smoking is prohibited, which may reduce harm by reducing smoking intensity, but could instead represent an increase in tobacco consumption among smokers.
Smokers generally view SLT as a less harmful alternative to cigarettes11; however, SLT may be used alone or with cigarettes. When used alone, SLT has been independently linked to major health risks, although these could vary depending on the product types. For instance, the levels of tobacco specific nitrosamine (TSNAs), which are established carcinogens, vary among different products.12 Harmful effects include oral and esophageal cancer,12,13 stomach cancer,14 pancreatic cancer,15,16 and cardiovascular diseases.17–20 Although less is known about the health effects of SLT when used jointly with cigarettes,21 dual use may lead to additional harmful effects by deterring smoking cessation. For instance, smokers may use these products as an alternative source of nicotine when they are not permitted to smoke, for example, due to clean air laws6,9,10,22–25 or may use these products as nicotine replacements when they are attempting to quit. Both of these circumstances have important implications for tobacco control. Thus, continued monitoring to understand trends in SLT product use is critical to inform national efforts to reduce the overall public health harms of tobacco.26
In understanding the public health impact of SLT use, as well as for other nicotine delivery products, such as e-cigarettes, it will be important to develop measures of use that can inform the transitions to sole and dual use of cigarettes.27 Unfortunately, the definition of SLT use has been inconsistent across national tobacco use surveys, making the characterization of prevalence trends challenging.3,28 In particular, the National Health Interview Survey (NHIS), the National Survey on Drug Use & Health (NSDHS), and the Tobacco Use Supplement of the Current Population Survey (TUS-CPS), collect information on SLT use differently.3,29 The NHIS defines current SLT use as every day or some days use of a SLT product.30,31 The NSDHS defines current SLT use as past 30-days use of a SLT product.27 Here, we explore patterns of SLT use under different definitions using the TUS-CPS survey. One of the goals of this study is to assess the sensitivity of SLT use trend estimates to the use definition.
Previously, Mumford et al.3 and Nelson et al.4 have described trends in SLT use under different definitions from the 1980s to 2000. In this article, we update these analyses with three additional waves from the TUS-CPS from 2003 to 2011, focusing on the relationship of SLT use and smoking prevalence. We also consider the relationship of SLT use to tobacco sales using data from the Federal Trade Commission Smokeless Tobacco Report. Currently, there is limited literature on the burden and determinants of SLT use in the United States. In addition to investigating the time trends of SLT use, we also examine the factors associated with SLT prevalence in the United States.
We used data from nine waves of the TUS-CPS, a series of nationally representative cross-sectional surveys: 1992/93, 1995/96, 1998, 1999, 2000, 2001/02, 2003, 2007, and 2010/11. The TUS-CPS collects national and state level representative data on tobacco use in the US household population.32 The survey includes a civilian, non-institutionalized population of age 15 years or older. Primary data collection was conducted by telephone interviews, but about 30% of interviews were conducted in-person in the household. In this study, we restricted the sample to self-respondents of age 18 or older.
In addition to prevalence analysis, we analyzed trends in SLT consumption from 1985 to 2011 using data from the 2011 Federal Trade Commission Smokeless Tobacco Report33 to examine if the trends in sales reflect the prevalence trends. SLT consumption per capita is calculated by dividing all SLT sales in pounds (in a particular year) by the corresponding total US population.
Prevalence of SLT use was calculated using aggregates of monthly samples for each survey wave. The wording of SLT use questions in TUS-CPS varies over time. Supplementary Table 1 summarizes the definitions of current use of SLT in TUS-CPS by survey year. For 1992/93 and 1995/96 surveys, current SLT use was defined as those reporting ever using SLT on a “regular basis” and answering “yes” to the current use question. Because the screening questions changed (removal of “on regular basis”), the baseline measure of current use for 1998 to 2010/11 surveys was defined as those reporting having ever used, responding “yes” to every day or some days current use, and reporting using SLT products at least 10 out of the past 30 days (as a criteria for “regular use”). We also explored the sensitivity of our results to varying measures of SLT use based on frequency from 1998 to 2010/11 (1, 20 and 25 days out of the past 30 days), and compared the SLT consumption per capita data with different definitions of SLT use.
Current cigarette smokers were defined as those who reported ever smoking at least 100 cigarettes during their lifetime, and reporting smoking every day or some days at the time of the survey. Current smokers were also distinguished by those reporting smoking every day versus some days, and those reporting quitting attempts within the past year. Former smokers were defined as those who reported ever smoking at least 100 cigarettes during their lifetime, and did not smoke at the time of the survey. Consistently with Mumford et al.,3 dual users were defined as those who are simultaneously classified as current smokers and current SLT users according to the definitions above, the later assuming the 10-day metric.
We categorized age into seven groups: 18–24 years, 25–34 years, 35–44 years, 45–54 years, 55–64 years, 65–74 years, and 75 years and older. Due to a low sample size for non-white racial/ethnic groups, race/ethnicity was collapsed into a dichotomous variable, white and Other race. Region was categorized into four categories: Northeast, Midwest, South and West. Education level was grouped into four categories: less than a high school degree, a high school degree or equivalent, some college, and at least a 4-year college degree. Household income was grouped into four categories: less than $14 999, $15 000–$34 999, $35 000–$74 999, and $75 000 or more. Residence was defined by the CPS’s use of Metropolitan Statistical Areas (MSAs), which divides major metropolitan and rural areas.
Prevalence was estimated using SAS 9.3 (SAS Institute, Cary, NC), applying sample weights and the PROC SURVEY procedures, which allow for adjustment of survey design. Trends in cigarette smoking, SLT use, and dual use prevalence (using both cigarette and SLT products) with 95% confidence intervals (CIs) for different demographic groups were calculated. To characterize trends in SLT prevalence, frequency (number of days used during the last 30 days), and SLT consumption, we performed joinpoint regression analyses using the statistical software Joinpoint, version 4.1.5 (Surveillance Research Program, US National Cancer Institute). Joinpoint regression identifies statistically significant trend change points and the annual percent change (APC) in each significant trend segment.34 Because SLT prevalence among females is low (less than 0.05% each year), examining the distribution of female SLT use by demographic characteristics yields unstable estimates; therefore, estimations were confined to adult males only.
Binary logistic regression analysis was used to identify potential demographic factors associated with SLT use. Only potential variables with a P value of less than .05 in univariate models were retained in the final multivariable logistic regression model. For the multivariate analyses, we categorized age into four groups: 18–24 years, 25–44 years, 45–64 years, and ≥65 years. The Cochran-Armitage trend test was performed to test for trends in age, education and income. For this analysis, odds ratios (ORs) with their corresponding 95% CIs are reported. Regression analyses were performed in SAS 9.3 using the PROC SURVEYLOGISTIC procedure.
Tobacco use prevalence for the total population, adult males and females (ages 18 and older), and males ages 18–24 is presented in Table 1. Smoking prevalence significantly decreased for adult males with APC rate of 2.7% per year (1998–2011) and for adult females with APC rate of 2.5% per year (1992–2011). For males ages 18–24, smoking prevalence remained approximately constant (nonstatistically significant increase of 0.6% per year) from 1992 to 2000 and then decreased at 4.1% per year from 2000 to 2011. Declines in overall SLT prevalence from 1992 to 2003 were observed assuming our baseline SLT use definition (frequency of at least 10 of the past 30 days) from 1998 to 2011. Total adult SLT use significantly decreased at an APC rate of 4.5% per year from 1992 to 2003. The SLT use prevalence has remained approximately constant since 2003 (nonstatistically different from zero). Similar patterns were also found in adult males (APC = −4.4%, 1992–2003) and males age 18–24 (APC = −9.5%, 1992–2003).
The dual use of SLT and smoking also declined. Dual use by adult males declined 5.3% per year, and by adult females decreased at about 8.4% per year throughout the period of analysis.
Table 2 shows SLT prevalence and trends among adult males by age, race, residence, education and current smoking frequency. Age-specific SLT prevalence is the highest for males ages 18–24 in early years (1992 and 1995 surveys), but the peak shifted to older ages since 1998 (highest for ages 25–34 from 1998 to 2003 and for ages 35–44 in 2007 and 2010/11). This shift may reflect trends in use by birth-cohort, with lower overall consumption of tobacco products in more recent cohorts. The SLT prevalence is highest among whites, those living in rural areas, those with lower education, and non-daily smokers across all survey years. The youngest age group (18–24) had the greatest decline in SLT prevalence from 1992 to 2003 (APC = −9.5%) compared with other age groups; however, the decrease in this and other groups stopped in 2003. Among race/ethnicity, males of other races had a larger decline in prevalence than whites (APCs = −10.6% vs −3.9%, respectively), although their rate decrease stopped earlier. Supplementary Table 2 shows the additional demographic characteristics of SLT use. Among income categories, those earning less than $14 999 per year had the largest significant decline in SLT prevalence compared with other groups (APC = −6.3%), although the decrease stopped in 1993. Among region, SLT prevalence decreased in all regions, with those living in the South having the largest significant declines in SLT prevalence (APC = −5.6%, 1992–2003) compared with other regions. Former and never smokers had also significant declines in SLT use (APCs = −2.4%, 1992–2011 and −5.2%, 1992–2003, respectively).
Trends in SLT consumption per capita from 1985 to 2011 are presented in Figure 1a. From 1985 to 1988, there was a significant 3.04% APC decrease per year in SLT consumption per capita. SLT consumption significantly decreased at 2.24% APC from 1991 to 1999, but has since decreased at 0.35% APC (1999–2011).
Starting in 1998, respondents were asked about their frequency of SLT use if they reported current use of SLT products every day or some days. Following Mumford et al., we defined, as our baseline metric, current regular users of SLT products as those reporting a frequency of at least 10 of the past 30 days. Among those, the average frequency of SLT use (number of days used during the past 30 days) did not vary in total adults, adult males and adult females, throughout the survey years (Table 1). Otherwise, the average frequency of SLT use declined in young males by about 1.7% per year from 1998 to 2002.
We also examined different thresholds for defining frequency of SLT use (at least 1 day, 20 days or 25 days of the past 30 days). Figure 1b shows the trends of different SLT prevalence assuming different frequency use thresholds (1-day, 10-day, 20-day, and 25-day) in adult males. Prevalence for all demographic groups for all definitions is shown in the appendix (Supplementary Table 3). The 1-day measure leads to 12%–20% (about 17% on average) higher SLT prevalence relative to the baseline (10-day) measure. However, the 1-day definition shows a uniform decline over time, while the 10-day, 20-day and 25-day show that independently of the definition; SLT prevalence decreased until 2003 and has been roughly constant since 2003. While the 10-, the 20-, and 25-day measures show similar patterns, the 10-day definition shows a less pronounced decline. We compared the SLT consumption per capita (Figure 1a) with the prevalence of SLT across definitions (Figure 1b), and consistent with previous results by Mumford et al., the SLT use trends, using the 10-day metric, more closely mirror the SLT consumption per capita compared with other definitions.
Due to low SLT prevalence among females, we only included males in the logistic regression analyses. In univariate analyses (Table 3), the likelihood of use of SLT decreases with increasing age (P trend < .0001). The likelihood of SLT use decreases with advancing education (P trend < .0001). People who live in the non-MSA areas (OR = 3.2, 95% CI: 3.0–3.4) are more likely to use SLT, residents of the South (OR = 2.8, 95% CI: 2.5–3.1), former smokers are likely to use SLT compared to never smokers, with an unadjusted OR = 1.7 (95% CI: 1.6–1.9). In a multivariable logistic regression model, ages 25 to 44 (OR = 1.3, 95% CI: 1.1–1.5), white race (OR = 3.2, 95% CI: 2.7–3.8), living in a non-MSA area (OR = 2.6, 95% CI: 2.4–2.8), unemployment (OR = 1.4, 95% CI: 1.1–1.7), lower education, for example, those with a high school degree or equivalent (OR = 1.4, 95% CI: 1.2–1.6), and being a former smoker (OR = 2.0, 95% CI: 1.8–2.2) were associated with SLT use after adjusting for other factors.
We examined trends in SLT and cigarette smoking use in the United States over a 20-year period using nine cross-sectional, nationally representative surveys. For US adults, significant declines in SLT prevalence for both genders were observed, particularly in males of ages 18 to 24 years old with a decrease of 9.5% per year from 1992 to 2003. However, we found that the decreases in SLT prevalence in most groups stopped since 2003. In contrast, the frequency of SLT use among current users remained approximately constant throughout the period of analysis at an average of 26 out of the past 30 days. Consistent with prevalence trends, SLT per capita consumption declined at a constant rate until 1999, but there was a significant reduction in the rates (APC) since 1999. In contrast, smoking prevalence showed a greater decline since 1998 than prior to that year. The information about SLT use in the TUS-CPS surveys has changed over time. In particular, the screening SLT use question was relaxed from “regular use” prior to 1998 to “at least one time” in more recent years (Supplementary Table 1), which may artificially increase the prevalence estimates for recent years relative to earlier years by omitting non-regular users who would have answered yes to ever use. To address this limitation, following the approach by Mumford et al.,3 we constructed different measures of SLT prevalence based on the reported use frequency, information that has been collected since 1998. Previous analysis of SLT3 and recent analyses of e-cigarettes by Amato et al.27 demonstrate the importance of accounting for use frequency when estimating tobacco products use prevalence. Here, we found that although the absolute prevalence estimates vary across definitions, with lower prevalence for higher frequency requirements (Figure 1b), the estimated male SLT prevalence trends show a consistent pattern of decline until 2003 and stability through 2010 (Figure 1b, Supplementary Table 3). The 20- and 25-day measures also show a flattening trend since 2003, but more rapid decline over 1992–2003 compared to the 10-day measure, suggesting that the 20- and 25-day measures are more sensitive metrics of changes in trend.28 The results also differ somewhat for male ages 18 to 24 compared with adult males, suggesting greater sensitivity to secular trends. Similarly to Amato et al.,27 we found that any past 30-day use does not necessarily detect changes in use trends.
Younger age, white race, unemployment, residents of the South, residents of rural areas lower education and former smoker status were found to be associated with SLT use. Those reporting less than 12 years of education were twice as likely to use SLT in comparison with those with at least a college degree. White males were nearly 3 times more likely to use SLT compared with those who are other races, thus showing the importance of race and SES. These findings are consistent with recent results by Bhattacharyya et al.,35 whose analyses found that whites are 2.5 times higher for active snuff use and 2.2 times higher for active chewing tobacco use, as well as with the earlier analyses of TUS-CPS data from 1992 to 2002 by Mumford et al.3
The Healthy People 2020 target36 is to reduce SLT prevalence for U.S adults to 0.3% or below. However, current prevalence is about 1.3% and at current trends, the target will likely not be achieved. The change in the decreasing trend of SLT use since 2003 may be partly attributable to the changing SLT landscape, with the introduction of new alternative nicotine delivery products in recent years, like snus37,38 and electronic cigarettes, which have reenergized the market, and the increasing influence of products with flavoring and portion pouch packaging.39,40 Previous studies have found important reductions in chewing tobacco sales, which may explain the decreases observed in SLT prevalence prior to 2003.39,41 In contrast, the promotion of moist snuff, snus and dissolvable tobacco, with “spit-free” formulations, may be responsible for the trend changes in SLT use.37 In addition, data from the CDC state system42 shows that cigarette taxes increased in many states from 2003 to 2010, while SLT taxes changed in fewer states and often by small amounts, which may explain the relative increase in SLT use compared to cigarettes (ie, the flattening in trend in SLT use while cigarette use declined more rapidly than in prior years).
Understanding the relative impacts of different products in shaping SLT consumption is needed. It is thus important to conduct studies of SLT use distinguishing by product category and use patterns, although the constant changes in the market and product category distribution, as well as changes in survey questions, make these studies challenging. Preliminary results (data not shown) using TUS-CPS suggest a progressive replacement of chewing tobacco by snus and new forms of SLT.
A strength of our study is having access to detailed tobacco use data from nationally representative surveys, which allows for quantitative joinpoint analyses of SLT use trends for the past two decades in relevant demographic groups. In addition, we examined SLT use trends and SLT consumption, allowing for comparisons between self-reported data and market sales. We were able to go beyond recent studies3,30 to examine trends in SLT use since 2003, and how patterns varied depending on the measure of use (number of days used in the past 30 days) and the sociodemographic group. Nonetheless, our findings should be interpreted with caution due to the following limitations.
First, while the TUS-CPS is a nationally representative survey, all information is self-reported, which may lead to underestimation of the true prevalence. Second, the survey data is cross-sectional, making it difficult to assess trends in initiation and cessation rates of SLT use. Third, the changes across survey years of SLT use definitions may have introduced bias, in particular going from the regular use to ever use screening question. Nevertheless, as discussed, we use information on frequency to attempt to increase the current use definition across years, and sensitivity analyses show that our general conclusions are consistent across alternative SLT use definitions. Lastly, tobacco companies have continued to introduce new SLT products on the market, making the interpretation of trends difficult given the variability in available products each year.
In sum, we report a significant declining trend in SLT product use from 1992 to 2003, suggesting the impact of tobacco control policies in reducing tobacco consumption. However, the decline ended in 2003 for SLT prevalence and ended in 1999 for per capita, highlighting the need for additional tobacco control efforts focusing on alternative tobacco products. These results are consistent with a recent analysis by Nguyen et al.43 that shows little change in SLT prevalence between 2011 and 2013 in most states, and increases in prevalence in some states such as Louisiana, Montana, South Carolina, and West Virginia. With the Healthy People 2020 goal, extra efforts are necessary in the next 5 years to achieve the 0.3% target. Focusing tobacco control efforts, particularly targeting at-risk demographic groups, is needed to further reduce SLT consumption in the United States. Little is known about whether the increasing use of e-cigarettes44,45 and the potential use of snus will impact future SLT trends. E-cigarettes may replace SLT use, especially snus, if it is viewed as a better harm reduction alternative. Continued monitoring of SLT use and its relationship to e-cigarette use is needed.
This work was supported by National Institute of Health/National Institute on Drug Abuse grant R01DA036497.