Changing age-specific patterns of cigarette consumption in the United States, 1992–2002: Association with smoke-free homes and state-level tobacco control activity

Introduction

In the early 1990s, the key smoking questions in national surveys in the United States were changed to better capture the cross-sectional prevalence of light or intermittent smoking behavior (Giovino et al., 1994). Although cigarette consumption levels below 5 cigarettes/day (CPD) have been labeled as insufficient for nicotine regulation and physiological dependence (Benowitz, Jacob, Kozlowski, & Yu, 1986), some smokers may maintain stable behavior at these low levels because of psychological dependence. Alternatively, for heavier smokers, a period of light or intermittent smoking may be part of a dynamic process of attempting to quit (Zhu, Sun, Hawkins, Pierce, & Cummins, 2003).

Using a longitudinal population survey, Zhu et al. (2003) observed that the majority of light smokers (defined as intermittent or daily and smoking less than 5 CPD) identified in California in the early 1990s did not maintain a stable consumption level over a 20-month period. The changes in consumption observed in these longitudinal data could be represented by a dynamic model of quitting and relapse, in which heavier smokers reduced smoking prior to making a quit attempt (Farkas, 1999) and relapse among former smokers began with a period of light smoking. However, the population trend was an increasing proportion of light smokers while overall smoking prevalence declined. This model fits the experimental work of Cinciripini, Hecht, Henningfield, Manley, and Kramer (1997), who have demonstrated that a structured reduction phase prior to a quit attempt can enhance successful quitting.

However, approximately a quarter of Californian light smokers in the early 1990s did maintain a stable pattern of consumption (Zhu et al., 2003). In a series of studies, Shiffman and colleagues have studied “chippers” who have maintained a light or intermittent smoking pattern for at least 2 years. Chippers, who smoked fewer than 5 CPD but more than 4 days/week, did not exhibit the characteristic features of nicotine dependence (Shiffman, Paty, Gnys, Kassel, & Elash, 1995), regardless of whether they had previously been heavier smokers (Shiffman, Paty, Gnys, & Zettler-Segal, 1994). In general, chippers were not social smokers as defined by Moran, Wechsler, and Rigotti (2004) as they smoked more when alone than with others (Shiffman & Paty, 2006) but rather had a smoking pattern similar to smokers who had started smoking only recently (McKennell & Thomas, 1967; Russell, 1971) with smoking behavior cued by stimuli such as drinking alcohol, eating, relaxing, and socializing (indulgent behaviors) as well as being strongly associated with negative affect (Shiffman & Paty, 2006). This pattern was markedly different from that of heavier smokers, in whom smoking behavior appeared to be cued by a need to maintain blood nicotine levels so as to avoid lapsing into nicotine withdrawal (Shiffman et al., 2002). Shiffman and Paty (2006) hypothesized that these chippers may have somehow been “fossilized” in an early stage of the process of becoming a smoker.

Increased prevalence of such fossilized smoking initiation might be associated with the introduction of restrictions on smoking (Pierce et al., 1991), especially as chippers tend not to smoke when working (Shiffman & Paty, 2006), and in population studies, the prevalence of intermittent smoking is associated with the prevalence of workplace smoking restrictions in the smoker's resident state (Tauras, 2004). We hypothesized that light and intermittent smoking behavior would have increased among young people along with the known increases of smoke-free workplaces (Gerlach, Shopland, Hartman, Gibson, & Pechacek, 1997) and smoke-free homes (Gilpin, White, Farkas, & Pierce, 1999) across the 1990s. Given that consumption by smokers is strongly influenced by race/ethnicity, with differences by age and acculturation in minorities (Evans et al., 1992; Shiffman et al., 2002; U.S. Department of Health and Human Services [USDHHS], 1998), to avoid confounding, we tested this hypothesis among the national non-Hispanic White population using large population surveys from 1992 through 2002.

Method

Results

The prevalence of current smoking decreased over time among all age groups over the 10-year study period: 1992–1993 to 2001–2002 (Figure 1). However, the drop in smoking prevalence was much smaller (p < .001) among those aged 18–29 years (1.2 percentage points) compared with those aged 30–44 years (3.6 percentage points) or those aged 45–64 years (3.8 percentage points). Over the same period, average consumption among continuing smokers declined in all age groups by more than 2 CPD: from 15.1 (95% CI = 14.9–15.2) to 12.5 (95% CI = 12.3–12.8; 2.6 CPD decline) in those aged 18–29 years, from 18.8 (95% CI = 18.7–19.0) to 16.2 (95% CI = 15.9–16.4; 2.6 CPD decline) among those aged 30–44 years, and from 20.6 (95% CI = 20.4–20.8) to 18.5 (95% CI = 18.3–18.8; 2.1 CPD decline) among those aged 45–64 years. This decline came, in part, from a consistent large decline (∼4 percentage points) in the prevalence of moderate-heavy smoking (at least 15 CPD) in each age group (see Figure 1).

A change in the prevalence of light smoking also contributed to the decline in average consumption among continuing smokers. Over the decade, the prevalence of smoking fewer than 5 CPD increased significantly (p < .001) among those aged 18–29 years (from 4.7% to 6.0%) so that 22% of current smokers in 2001–2002 in this age group were light or intermittent smokers (see Figure 1). This increase was significantly greater than for either of the older age groups (p < .03). At the end of the decade, 4.1% of the 30- to 44-year-old age group (17% of current smokers) and 2.5% of the 45- to 64-year-old age group (12% of current smokers) were in this lightest consumption category, up from 3.5% and 2.4%, respectively, in the 1992–1993 survey.

Discussion

Throughout the 1990s, the average consumption level among smokers decreased markedly in the non-Hispanic White majority population of the United States. A major decline in the population prevalence of smokers with high consumption levels (at least 15 CPD) was observed in all age groups. This decline was associated with an equivalent decline in overall smoking prevalence in populations aged more than 30 years. Among young adults (aged 18–29 years), the decline in overall smoking prevalence was considerably smaller, and it appeared to be offset by an increase in the population prevalence of light smokers (those consuming fewer than 5 CPD, of whom 88% reported smoking on fewer than 7 days/week). Given the known association between cigarette consumption and lung cancer (Burns, 1997; Doll & Peto, 1978; Flanders, Lally, Zhu, Henley, & Thun, 2003), this change in population smoking patterns will translate into a significant reduction in smoking-related illnesses in the future.

Over the decade, the proportion of smokers who had smoke-free homes more than doubled in each age group. This increase was strongly associated with the change in consumption patterns across the nation. As has been noted previously, given the large amount of time a person typically spends at home, having a smoke-free home raises a significant barrier to opportunities to smoke (Biener, Cullen, Di, & Hammond, 1997). Previous research suggests that having a smoke-free home may be a marker of social pressure against smoking in the immediate social network (Gilpin et al., 1999). Living with children or nonsmoking adults increased the probability of a smoke-free home, and this probability was increased significantly in the presence of a family preference that the smoker successfully quit.

These barriers to smoking also may explain the higher prevalence of light and intermittent smoking in young people with smoke-free homes. Young people who have grown up with a smoke-free home, school, and workplace environment may stabilize at a much lower dependence level than those without such restrictions. Indeed, among young adult smokers, the increase in the proportion of light smokers over time was mediated by the increase in smoke-free homes to the extent that a secular time trend was not significant in our multivariate models.

In our model, young adult smokers who lived in states with stronger tobacco control activities were more likely to be light smokers. This effect was independent of the presence of a smoke-free home and increased with increasing state level of tobacco control activity, suggesting that the social norms around smoking also have an independent impact on the general consumption pattern. Educational level also was independently associated with light smoking. Indeed, we found an interaction effect between education level and the presence of a smoke-free home. Among smokers without a college education, the relative effect of a home smoking ban was stronger, associated with a one-third greater increase in the proportion of light smokers, although the absolute levels of light smoking remained higher among the more educated. If the observed associations are causal, promotion of smoke-free homes may be an effective tobacco control strategy to reduce educational disparities in tobacco use.

As national cessation rates have been rising (Messer et al., 2007) and as low-rate smoking has been postulated to be an integral part of the quitting process (Zhu et al., 2003), it is not surprising that more than 10% of smokers now fall into this light or intermittent consumption category for each age group. However, the strong age gradient observed in the increase of light or intermittent smoking suggests that a change in the pattern of initiation may be occurring as well, as suggested previously (Shiffman & Paty, 2006). Further supporting a change in initiation, young adult smokers were almost twice as likely to be light or intermittent smokers compared with adults aged 50–64 years, and this level of difference between the two age groups is not seen in quitting behavior (Messer et al., 2007).

Many of these light or intermittent smokers had never smoked daily, and this was particularly the case for young adult smokers. Also, in our multivariate models, the increase in smoke-free homes was associated with the increase in light consumption among young adult smokers. These observations are consistent with the hypothesis that smoke-free homes would interrupt the smoking initiation process among young adults, preventing their progression to heavier smoking levels and thus contributing to the increase in light smoking in this age group. However, among light smokers, the proportion who had never smoked daily did not change over the study period, which was unexpected. Our data also are consistent with the hypothesis that lighter smokers will more readily adopt a smoke-free home and, thus, that the higher prevalence of smoke-free homes among young adult smokers is more a consequence than a cause of their higher rates of light smoking. A future analysis of these cross-sectional data by birth cohort may help to clarify these questions.

It is also possible that tobacco industry marketing contributed to the increase in light smoking among young adults. Had smoking initiation rates been lower, the prevalence of light smoking would have been lower among young adults, regardless of whether they later progressed to heavier smoking. This birth cohort of young adults was previously shown to be strongly receptive to tobacco industry marketing influences in the early 1990s (Evans, Farkas, Gilpin, Berry, & Pierce, 1995). Further, these young adults were exposed to the increased tobacco industry marketing in bars and clubs that targeted this cohort around the turn of the 21st century (Gilpin, White, & Pierce, 2005). It is possible that receptivity to tobacco marketing as a young teenager may predict increased susceptibility to tobacco marketing later in life, such as the young adult years. Thus, the tobacco industry may have at least partly succeeded in maintaining smoking initiation rates among young adults even in the face of increasing smoke-free homes and state-level tobacco control activities. Unfortunately, we are limited in exploring this area further as the TUS-CPS does not include any questions on receptivity to tobacco industry marketing.

A major limitation of the present study is that it focused on the majority non-Hispanic White U.S. population to avoid confounding by the known major ethnic differences in smoking consumption patterns. Future work should investigate these questions among Hispanics and other non-White populations. This study is a national population survey enabling estimates of population behavior, and it is limited by the response rate, as are all such surveys. However, compared with other population surveys conducted over the time period, response rates in the TUS-CPS were among the highest at 61%–68%. Further, the large declines in response rates seen in telephone-based surveys (Biener, Garrett, Gilpin, Roman, & Currivan, 2004) were not a major factor in this survey. Although the smoking measure was self-reported and not biochemically validated, considerable evidence indicates that population surveys provide an unbiased estimate of smoking behavior (Gilpin et al., 1994; Pierce et al., 1987). An additional limitation of the present study is that the last publicly available survey in this surveillance series is 2001–2002. Finally, our projection of future health benefits does not factor in any compensation that may occur among smokers. In experimental settings, smokers are known to increase the intensity of their smoking when their access to cigarettes is limited (USDHHS, 1988).

In summary, population increases in both smoke-free homes and social norms about smoking have shown a strong association with reduced consumption levels among smokers, both over time and across states. This pattern change should translate into a reduction in the population health risk for smoking-related diseases in the near term and represents a major success story for national tobacco control programs. However, the absolute increase in the prevalence of light or intermittent smoking among young adults in recent years is troubling. Whether this increase indicates the effectiveness of tobacco industry marketing efforts, particularly in maintaining smoking initiation, warrants further investigation.

Funding

Tobacco Related Disease Research Program grant (15RT-0238) from the University of California.

Declaration of Interests

None declared.

Supplementary Material