Tobacco use is the leading cause of preventable morbidity and mortality among adults in the United States and worldwide (
1–
3). Tobacco use causes 480,000 (i.e., 1 in 5) deaths per year in the United States (
4). Tobacco use has been associated with numerous health consequences, including coronary heart disease, stroke, chronic obstructive pulmonary disease, lung and other cancers, and increased risk of preterm delivery and low birth weight (
1). Tobacco use can result in nicotine dependence, which is highly comorbid with alcohol and drug use disorders as well as mood, anxiety, and personality disorders (
5,
6). Tobacco use also imposes enormous economic costs in the United States, recently estimated at $193 billion per year (
7).
Prevalence of cigarette use in the U.S. general population has decreased markedly since the 1960s, but several trend studies have shown smaller declines in more recent decades, suggesting that large population improvements in cigarette use have lessened (
8–
10). Although tracking changes in cigarette use is common, less is known about changes in overall tobacco product (i.e., nicotine) use or nicotine dependence or other proxy measures used to characterize severely dependent users (
11). The “hardening hypothesis” posits that as tobacco use declines, less dependent users will quit, leaving a growing proportion of severely dependent users who may be less likely to quit, resulting in a leveling off of declines in smoking (
12,
13).
Studies examining the hardening hypothesis have generally focused on proxy measures of nicotine dependence such as increased quit attempts, decreased ability to abstain, successful abstention, number of cigarettes per day, and time to first cigarette within 30 minutes of awakening or a combination of these measures. Results from these studies, conducted largely outside the United States between 2009 and 2015, are mixed, with some supporting (
14,
15) and others rejecting the hardening hypothesis (
16–
19). Other studies used cotinine levels as a proxy measure of nicotine dependence and found that levels had not changed between 1988 and 2012 (
20).
Results
Nicotine Use
Although the 12-month prevalences of nicotine use declined from 27.7% in the 2001–2002 survey to 26.9% in the 2012–2013 survey, when adjusted for sociodemographic characteristics, the adjusted risk difference (1.4%) denoted a small but significant increase in these estimates (
Table 1). Increases in 12-month nicotine use were also seen among men (adjusted risk difference=1.3%), women (adjusted risk difference=1.4%), 30- to 44-year-olds (adjusted risk difference=2.8%), those age 65 or older (adjusted risk difference=1.2%), whites (adjusted risk difference=1.3%), blacks (adjusted risk difference=2.5%), the previously married or never married (adjusted risk differences, 1.8% and 1.67%, respectively), those with high school or less education (adjusted risk differences, 4.0% and 2.6%, respectively), those with incomes less than $35,000 or less than $20,000 (adjusted risk differences, 5.0% and 1.9%, respectively), and those residing in urban areas and in the Midwest and South (adjusted risk differences, 1.3%, 1.8%, and 1.7%, respectively).
In analyses testing whether changes in prevalence across surveys differed by sociodemographic strata (adjusted additive interaction p values), adjusted risk differences in nicotine use were significantly greater among those with high school or less education (4.0% and 2.6%, respectively) relative to those with some college or higher education (0.2%), and among those with incomes less than $69,999 compared with those of $70,000 or more.
Nicotine Dependence
The prevalence of nicotine dependence increased between the 2001–2002 (12.8%) and 2012–2013 (14.0%) surveys (
Table 2). Significant increases were seen among nearly all sociodemographic subgroups, except 18- to 29-year-olds, Native Americans, Asians/Pacific Islanders, and respondents with incomes of $70,000 of more. The adjusted risk difference of nicotine dependence was greater among men (3.4%) than women (1.9%); among 30- to 44-year-olds (3.5%) than those age 65 or older (1.8%); among the previously married (3.7%) compared with married respondents (2.2%); among those with high school or lower education (5.5% and 4.3%, respectively) compared with those with some college or higher education (1.1%); and among those in lower income groups (2.4%−5.6%) relative to those with incomes of $70,000 or more (0.1%).
DSM-IV Nicotine Dependence Among Users
Between the 2001–2002 and 2012–2013 surveys, the prevalence of 12-month nicotine dependence among 12-month nicotine users increased from 46.1% to 52.0% (
Table 3). In the adjusted analyses, increases were seen in nearly all sociodemographic subgroups except Native Americans, Asians/Pacific Islanders, respondents with incomes of $70,000 or more, and respondents residing in the Midwest. The adjusted risk difference of nicotine dependence among users was greater among men (8.0%) than women (4.6%); among Hispanics (12.5%) than whites (5.6%); among respondents with high school or less education (8.8% and 9.2%, respectively) than those with some college or higher education (4.0%); and among the lowest income groups (9.1%) compared with those with incomes of $70,000 or more (3.8%). The adjusted risk difference was lower among all three younger age groups (4.2%−7.6%) than the oldest age group (14.0%) and among those residing in the Midwest (2.6%) than in the West (9.2%).
Approximate Fagerström Dependence
Overall, similar to DSM-IV nicotine dependence, the prevalence of approximate Fagerström dependence increased (adjusted risk difference=1.8%), with increases among many sociodemographic subgroups, except 18- to 29-year-olds, Native Americans, Asians/Pacific Islanders, respondents with incomes of $70,000 or more, those with some college or higher education, and those living in the Northeast or the West. Similar to DSM-IV nicotine dependence, the adjusted risk difference of approximate Fagerström dependence was greater among those with high school or lower education (3.3% and 4.2%, respectively) compared with those with some college or higher education (0.6%) and among those in lower income groups (1.5%−4.7%) relative to those with incomes of $70,000 or more (−0.5%). Among 12-month nicotine users, similar to DSM-IV nicotine dependence, the prevalence of approximate Fagerström dependence increased (adjusted risk difference=3.1%), with increases among men, individuals age 65 or older, whites, Hispanics, married, never married, high school education, lowest and middle income groups, urban location, and the South region. Different from DSM-IV nicotine dependence, adjusted risk differences for approximate Fagerström dependence did not differ significantly by sociodemographic characteristics. See Tables S1 and S2 in the online supplement for details.
Discussion
The overall unadjusted 12-month prevalence of nicotine use showed a modest decline over the period between the 2001–2002 NESARC and the 2012–2013 NESARC-III, consistent with declines documented in the National Survey on Drug Use and Health, which found that rates of use of “any tobacco product” declined from 32.2% to 26.9% between 2002 and 2012 (
41). However, with adjustment for sociodemographic characteristics, the 12-month rates of nicotine use significantly increased between the 2001–2002 and 2012–2013 surveys. These results are primarily due to increases in nicotine use among population subgroups, especially those disadvantaged in terms of education and income. These findings underscore the importance of adjustment in studies examining changes over time in nicotine use. Larger increases were seen in 12-month nicotine dependence and nicotine dependence among 12-month nicotine users. The modest increase in 12-month nicotine use seen in this study is consistent with slight increases or the slowing in the decline of rates in current cigarette use among adults in the United States during the same time period (
8,
10,
14,
15) and the prevalences of cigarette or nicotine use in the 2001–2002 and 2012–2013 periods seen in other national surveys (
8,
10,
14,
15,
41,
42). The absence of studies on DSM-IV-defined nicotine dependence preclude further comparisons with previous studies.
The smaller increase in 12-month nicotine use relative to the much larger increases in 12-month nicotine dependence among 12-month users seen in this study suggests that increases in nicotine dependence are largely the result of increases in nicotine dependence among nicotine users and not increases in the prevalence of nicotine use. Taken together, these results support the hardening hypothesis. These findings are at variance with some (
16–
18,
43,
44) but not all (
14,
16) studies examining changes over time using proxy measures of nicotine dependence (e.g., cigarettes/day, time to first cigarette, and quit attempts). Sensitivity analyses showing similar increases in the approximate Fagerström dependence measure in our sample supports the validity of the finding for DSM-IV nicotine dependence. Although tobacco control measures may encourage less dependent individuals to quit, such efforts may not be effective among individuals who are already nicotine dependent (
23). Since tobacco control measures have had a significant effect on reducing nicotine consumption, future studies on the hardening hypothesis should use measures of nicotine dependence that do not rely heavily on nicotine consumption (
11,
45,
46).
In line with previous research (
8–
10), men had greater prevalences of nicotine use than women during both survey periods. Further, adjusted risk differences in nicotine use, nicotine dependence, and nicotine dependence among users increased among both men and women, but changes in adjusted risk differences for nicotine dependence and nicotine dependence among users over the survey period were greater among men than women. In previous studies using consumption-based measures of nicotine dependence, severely dependent women smokers were more likely to have seen a health care provider and to have received cessation advice from a provider (
47). Health care provider visits have been shown to increase the probability of making a cessation attempt and achieving a successful cessation attempt (
48–
50). Severely dependent female smokers were also less likely than severely dependent male smokers to be exposed to smoking restrictions at home and work (
47). Despite greater increases in nicotine dependence and nicotine dependence among users among men, women experience unique nicotine-related conditions associated with pregnancy, prenatal outcomes, oral contraceptive use, cervical cancer, and osteoporosis (
7). Further research examining gender-specific outcomes in nicotine dependence treatment is warranted.
Results by age group suggest that nicotine dependence emerges in later years of nicotine use and may be persistent despite the awareness of or actual health concerns accompanying older age. Reasons for the greater increases among older individuals of nicotine dependence and nicotine dependence among users may include social and historical influences of smoking initiation, social isolation, perceived benefits of smoking (e.g., as a preventative for depression), and unrealistic optimism (
11,
51–
53). Research is needed on the characteristics and motivations of nicotine use and nicotine dependence among older adults, who are especially vulnerable to nicotine-related physical disabilities, that can help inform age-specific interventions.
Previous studies on the hardening hypothesis using consumption-based measures of nicotine dependence have not examined race/ethnicity differences in detail, although a few studies report that white race is correlated with persistent nicotine use (
11,
54,
55). Although reasons for increases in nicotine dependence and nicotine dependence among users among whites, blacks, and Hispanics are unclear, further study on the social, cultural, behavioral, and genetic factors influencing nicotine use and the development of nicotine dependence among race/ethnicity groups is warranted, with special attention to access to treatment as a key factor.
Across the surveys, increases in nicotine use, nicotine dependence, and nicotine dependence among users were especially notable among individuals with lower education and income, and changes over time in these nicotine outcomes were greater among those with lower socioeconomic status relative to those with greater incomes and education. These findings are consistent with other studies that have characterized lower socioeconomic status as a characteristic of dependent or severely dependent smokers (
16,
56–
58) and other research showing a decline in severely dependent smoking among higher-socioeconomic-status smokers but not among lower-socioeconomic-status smokers (
59). These socioeconomic status disparities in nicotine use and nicotine dependence may be attributed to social disadvantage, shared social environments with other nicotine users, and fewer limitations on nicotine use inside and outside the home (
59). It is also possible that less educated individuals have lower access to information on the health consequences of nicotine use and to treatment services, or may be less responsive to health promotion (
60,
61). A greater understanding of increases in nicotine use and nicotine dependence among individuals of lower socioeconomic status is needed, especially in light of the disproportionate prevalence of nicotine-related morbidity and mortality among in this population (
60).
This study has several limitations. Nicotine use was assessed by self-report, and future national surveys should include nicotine biomarkers. Social desirability bias may have contributed to underreporting, and thus our study may underestimate nicotine outcomes. The NESARC-III response rate was acceptable and was similar to other concurrent national surveys (60.1%) but was lower than that of the NESARC (81.0%). However, weighting facilitated comparisons between the surveys (
62). The employers of the NESARC and NESARC-III were different—the U.S. Census Bureau and Westat, respectively. However, both surveys were presented to respondents as voluntary, and both were conducted under the auspices of the federal government. In addition, respondents in the NESARC-III received a modest payment for their time and effort, whereas those in the NESARC were not paid. While such changes in survey methodology are unlikely to have influenced participants’ response propensity in answering questions about tobacco and nicotine use (i.e., the primary study outcomes), this (and other limitations) is a reminder that replication in other studies is warranted. The NESARC and NESARC-III provided assessments at two time points, and potential fluctuations during the intervening years cannot be determined. The NESARC-III but not the NESARC collected DNA for genetic analyses. However, saliva samples were collected after questionnaire administration, minimizing any differential impact on survey estimates presented here. Nicotine dependence was not ascertained for each nicotine-containing product, so product-specific analysis could not be conducted. The FTND was not included in either the NESARC or the NESARC-III data sets, but an approximate measure based on four of the six Fagerström questions was created, and it included the two questions often considered the “core” Fagerström questions (
63). Limitations are balanced by the numerous strengths of the NESARC and NESARC-III, including large sample sizes, rigorous methodology, reliable measures of nicotine use and nicotine dependence, and non-consumption-based measurement of nicotine dependence.
In summary, the results of this study support the hardening hypothesis. Because of this hardening, the remaining nicotine users may be less likely to quit because of dependence, and the allocation of nicotine intervention services may need to be reconsidered. Some individuals appear to be unlikely to quit with public health efforts (e.g., increased price, elimination of advertising, protections against secondhand smoke), so evidence-based treatment interventions should supplement these public health activities. Comprehensive population-based and treatment-oriented interventions hold great promise in reducing nicotine use and nicotine dependence as a devasting cause of preventable morbidity and mortality. This study identified notable increases in nicotine dependence and nicotine dependence among users in men, middle-aged and older individuals, whites, blacks, Hispanics, and the socioeconomically disadvantaged. These subgroups would benefit most from targeted nicotine dependence intervention programs to help them overcome their dependence and quit nicotine use.