• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of amjphAmerican Journal of Public Health Web SiteAmerican Public Health Association Web SiteSubmissionsSubscriptionsAbout Us
Am J Public Health. 2002 February; 92(2): 250–256.
PMCID: PMC1447052

Chewing Tobacco: Who Uses and Who Quits? Findings From NHANES III, 1988–1994


Objectives. This study identified mutually exclusive groups of men at high and low risk for use of chewing tobacco and for quitting.

Methods. Analyses used a national sample of 1340 non-Hispanic Black, 1358 Mexican American, and 1673 non-Hispanic White men, aged 25 to 64, who participated in the National Health and Nutrition Examination Survey III from 1988 to 1994. Signal detection analysis was used to delineate high- and low-risk subgroups; survival analysis was used to estimate hazard curves for comparing age at onset for chewing tobacco use with that for smoking.

Results. Rural, lower-income Black and White men had the highest regular use of chewing tobacco (33.3%), followed by rural, higher-income men regardless of race/ethnicity (14.9%). Southern men who began using chewing tobacco during adulthood had the lowest quit rate (22.5%). In sharp contrast to smoking, chewing tobacco showed a continued onset throughout adulthood.

Conclusions. Because subgroups of men show highly different chewing tobacco use and quit rates and because age at chewing tobacco onset occurs across the life span, prevention and cessation programs should be specific to different risk groups and distinct from smoking programs.

More than 5 million men and women in the United States use chewing tobacco, which contributes to serious diseases of the mouth and heart and several types of cancer.1–8 The prevalence of chewing tobacco use varies widely by region and sociodemographic factors. Rates are highest in the South and rural areas,8 and men use chewing tobacco at 10 times the rate for women.9 Among men, American Indians/Alaskan natives and Whites have the highest rates of chewing tobacco use: 7.8% and 6.8%, respectively. Black, Hispanic, and Asian and Pacific Islander men use at much lower rates: 3.1%, 1.5%, and 1.2%, respectively. Among women, Blacks and American Indians/Alaskan natives have the highest rates: 2.9% and 1.2%, respectively. White and Hispanic women have the lowest rates: 0.3% and 0.1%, respectively. Asian and Pacific Islander women almost never use chewing tobacco.9 Chewing tobacco use rates increase as years of education decrease for both men and women. For example, 7.7% of the men with less than a high school education use chewing tobacco compared with 2.5% of the men with a college education or more.8

Over the last 20 years, prevalence rates for most groups have decreased; only White men aged 18 to 34 showed a higher prevalence rate in 1991 than in 1970.9 Despite recent trends suggesting that the overall use of chewing tobacco may be stabilizing,9 the continuing danger is that prevalence rates may not decrease further or that they may increase. The tobacco industry has shown remarkable flexibility in adjusting its marketing strategy to keep tobacco use at high levels. For example, the prevalence of chewing tobacco use among White men aged 18 to 24 increased 5-fold (from 2.2% in 1970 to 10.1% in 1991) during an extensive chewing tobacco advertisement and promotional campaign targeting this specific age group.10–12 Furthermore, the tobacco industry and even some researchers have suggested that switching to chewing tobacco may be an acceptable, safer alternative to smoking; this suggestion can only serve to increase its popularity.13 These concerns underscore the need to identify patterns of use among different sociodemographic groups and those at highest risk.

National surveys on the health of Americans, such as the National Health Interview Survey and the Behavioral Risk Factor Survey, regularly report overall prevalence rates of chewing tobacco use. However, these reports rarely present in-depth information about the characteristics of those who have chewed regularly and those who have quit. In the current analyses, we were interested in the relation of multiple sociodemographic factors to chewing and quitting and whether these factors delineated distinct high- and low-risk subgroups. Such delineation can aid in the development of tailored prevention and cessation programs and can help counter the pro-tobacco influences that promote chewing tobacco use.

In this study, we used signal detection methodology to identify distinct subgroups of men who had ever used chewing tobacco regularly and who had quit, based on national data collected in the National Health and Nutrition Examination Survey III (NHANES III). In addition, we compared the age at onset for chewing tobacco users and for cigarette smokers to contribute to our understanding of the differences between each form of tobacco use. We focused our study on men because very few women use chewing tobacco.9


NHANES III was designed primarily to assess the health and nutrition status of the US noninstitutionalized civilian population of non-Hispanic Whites, non-Hispanic Blacks, and Mexican Americans.14 NHANES III oversampled the 2 largest racial/ethnic minority groups in the United States—Blacks and Mexican Americans—to produce statistically reliable results for both. The survey used a stratified, multistage probability cluster sampling, similar to that used by previous NHANES surveys.15,16 Weights indicating the probability of being sampled were assigned to each respondent, enabling results to represent the US population for each group. NHANES III was conducted at 89 locations from 1988 to 1994 and included a total sample of 33 994 persons, aged 2 months and older. Data were collected via standardized interviews in participants' homes, and most participants had medical examinations at mobile centers. Response rates were high; 78% completed both the home interview and the medical examination.

Our analytic sample included 1340 Black, 1358 Mexican American, and 1673 White men between 25 years (to ensure that most men had completed their education and passed the average age for initiation of chewing tobacco) and 64 years of age (to avoid problems of selection effects resulting from typical life changes at older ages, such as poorer health, retirement, and health care coverage).17 We excluded data for men from other racial/ethnic groups (n = 206 “other Hispanic” and “non-Hispanic”) and interviews that were coded as unreliable (n = 10). We do not present descriptive data on the characteristics of the NHANES III sample because they have been published elsewhere.17



Age was calculated in years from the participants' reported birthdate. Respondents who reported their race/ethnicity as nonHispanic Blacks, Mexican or Mexican Americans, or non-Hispanic Whites were included. Education was measured as the highest grade or year of school completed. Family income was the total combined family income during the past 12 months, and family income ratio was family income divided by family size. Urban status was defined as “central and fringe metropolitan counties of 1 million or more,” and other counties with fewer than 1 million were designated as rural status. Geographic region of residence was defined as Northeast, South, Midwest, and West.

Tobacco use.

Respondents were classified as lifetime regular users of chewing tobacco if they reported ever having used chewing tobacco or snuff (yes/no) and stated an age when chewing tobacco was first used fairly regularly. Current chewers reported using chewing tobacco at the time of the interview (yes/no). Quitters were classified as lifetime regular users not currently using chewing tobacco. Current smokers had smoked at least 100 cigarettes in their lifetime (yes/no) and currently smoked (yes/no). Age at chewing tobacco onset was defined as the age when chewing tobacco was first used fairly regularly. Age at cigarette smoking onset was defined similarly. Amount of chewing tobacco use was defined as the number of containers used each week.

Less than 1% of the data were missing for the variables we analyzed, except for family income, which had 8% of the data missing.

Data Analysis

Signal detection analysis.

We used signal detection analysis, a method of recursive partitioning (classification trees), to identify distinct risk subgroups based on the sociodemographic factors.18,19 This methodology is particularly well suited for identifying interrelationships between multiple factors related to poor health and groups at high and low risk for a particular health outcome. Signal detection models sequentially partition the data to identify subgroups that are mutually exclusive and maximally discriminated from each other, based on the probability of the outcome, which must be binary. The independent variables are entered, and the algorithm then selects a variable and cutpoint based on a combined optimal measure of sensitivity and specificity with regard to the outcome variable.

Unlike regression analysis, signal detection analysis is not highly affected by the collinearity of variables in the model (e.g., family income and family income ratio). After choosing and splitting on the first optimally efficient variable, the signal detection program separately searches each subgroup or “branch” of the first split for the next most efficient variable and cutpoint, again using all initial independent variables as candidates. This procedure is repeated separately in each subgroup and ends when (1) subgroup samples become small (n < 25), (2) no further significant discriminating variables are found (P < .001), or (3) no further independent variables remain. The results are displayed as classification trees.

We ran separate signal detection analyses for the 2 binary (yes/no) outcome variables of lifetime regular use of chewing tobacco and quitting. Each signal detection analysis included age, education, family income, family income ratio, race/ethnicity, urban or rural status, and geographic region of residence. Age at onset was included in the analyses of quitting.

Subgroup profiles.

We used SUDAAN, Version 7.11 (Research Triangle Institute, Research Triangle Park, NC), to calculate weighted means, percentages, and standard deviations to examine the characteristics, or profiles, of the subgroups identified by the signal detection analysis. SUDAAN incorporated sampling weights that adjusted for unequal sampling probabilities resulting from the complex NHANES III sample design. The subgroup profiles included all of the variables used in the signal detection models as well as amount of chewing tobacco used each week and smoking status.

Survival analysis.

We conducted survival analyses with SAS, Version 6.12 (SAS Institute Inc, Cary, NC), to estimate the hazard rates for age at onset for chewing tobacco use and cigarette smoking. The hazard curve for each event gives the instantaneous rate of the event at each time point. Unlike a survival curve, which can only decrease, hazard curves can increase and decrease depending on the underlying event probabilities, allowing one to observe more easily whether different periods of life are subject to higher risk. In this analysis, our interest was to create a graphical representation of the pattern of risk to compare the 2 types of tobacco use rather than to compare the actual incidence rates. The hazard plot was preferred over a simple bar graph of frequencies because of the large differences in the prevalence rates of chewing and smoking.


We present the results of the signal detection analysis for each outcome variable followed by a profile of the distinct subgroups identified by the analysis. High-risk and low-risk groups were defined in relation to the overall rate of the dependent variable.

Signal Detection Analysis and Subgroup Profiles

Lifetime regular use of chewing tobacco.

The first signal detection analysis examined factors associated with lifetime regular use of chewing tobacco. The signal detection model identified 5 subgroups (Figure 1 [triangle]). The overall rate of lifetime use was 10.7%, and high-risk and low-risk groups were defined in relation to this overall rate. The first sociodemographic variable that distinguished lifetime regular users was urban or rural status. Among rural men, family income and race/ethnicity distinguished 2 subgroups with significantly different rates of use. Black and White men with family incomes of $8500 or less reported the highest regular use rate (33.3%, group 1), whereas Mexican American men with family incomes of $8500 or less reported a substantially lower rate (7.7%, group 2). Regardless of race/ethnicity, rural men with family incomes greater than $8500 had a higher rate of regular use (14.9%, group 3) than men overall. Among urban men, geographic region of residence was the only additional distinguishing variable. Men living in the South had more than twice the rate of regular use (9.1%, group 5) as men not living in the South (4.4%, group 4), but rates for both groups were lower than the overall rate.

—Groups of men at high and low risk for lifetime regular use of chewing tobacco (10.7% overall rate, n = 4371).

Of the 5 subgroups identified by the signal detection analysis, rural, low-income Black and White men (group 1) constituted the highest risk subgroup. As shown in Table 1 [triangle], these men were predominantly Black (65%) and from the South (76%). Although their mean age was 44.9 years, most had not completed high school (62%), and their average family income was very low ($5594). Rural, higher-income men (group 3) were the next highest risk subgroup. These men were predominantly White (49%), with the remainder split equally among Blacks (26%) and Mexican Americans (25%). Two thirds of these men were from the South (62%). These men were more highly educated than the highest risk group, with 65% completing a high school education or higher, and their income was substantially higher ($32 905). Among those who were still current users of chewing tobacco, men in the 2 highest risk subgroups used more chewing tobacco each week than did men in the low-risk subgroups (4.1 or more containers per week vs 3.2 or fewer). Men in the highest risk group were much more likely to be smokers (61%) than were men in the other groups (35%–39%).

—Profile of Figure 1 [triangle] Groups at High and Low Risk for Lifetime Regular Use of Chewing Tobacco: NHANES III, 1988–1994

Quitting chewing tobacco.

Among the 10.7% of the men with a lifetime history of regular use (n = 466), 60.3% reported no current use and were classified as quitters (Figure 2 [triangle]). The signal detection analysis identified geographic region of residence as the first distinguishing variable, with nonsouthern men reporting higher rates of quitting (72.3%, group 1) than the overall rate. For southern men, age at onset was the next distinguishing variable and yielded 2 groups: those younger than 38 years and those 38 years and older. Men who started using chewing tobacco at younger ages were substantially more likely to quit than were men who started using chewing tobacco at older ages (58.5%, group 2; 22.5%, group 3).

—Groups of men who had been lifetime regular users of chewing tobacco at high and low risk for quitting (60.3% overall rate, n = 466).

The profile for the men who were the least likely to quit chewing (southern men who started after young adulthood, group 3) was distinctly different from that for the men in the other 2 groups. The age they first used chewing tobacco regularly (47.3 years) was more than 20 years older than the other groups (Table 2 [triangle]). A much larger proportion of these high-risk men had not completed high school (65% vs 26%, group 1; 43%, group 2), but they were less likely to be smokers (25%) compared with the other 2 groups of men (34%, group 1; 41%, group 2).

—Profile of Figure 2 [triangle] Groups at High and Low Risk for Not Quitting Use of Chewing Tobacco: NHANES III, 1988–1994

Survival Analysis

We compared the shape of the 2 hazard curves generated by the survival analysis: the age at onset for regular use of chewing tobacco and the age at onset for regular cigarette smoking (Figure 3 [triangle]). Both hazard curves showed a rapid increase in onset of tobacco use beginning at about age 10 and peaking at about age 18. For cigarettes, the decrease after the peak hazard was quite steep, but onset became negligible after age 35. In sharp contrast, for chewing tobacco, the decrease after the peak was more gradual than for smoking, and a second increase was evident in adulthood, reaching a peak at age 55. The hazard curve then decreased steeply, but the curve never became flat, suggesting a continuing risk for onset of chewing tobacco use late in life.

—Hazard curves for age at onset for regular cigarette smoking and chewing tobacco use for men aged 25 to 64 years.


In this study, we used signal detection analysis to identify mutually exclusive groups of men at high and low risk for lifetime regular use of chewing tobacco and for not quitting. Two high-risk groups for regular use of chewing tobacco were identified, and 1 high-risk group was identified for not quitting. These results identify distinct groups of men in need of chewing tobacco prevention and cessation programs. We also used survival analysis to compare the age at onset for chewing tobacco use with the age at onset for cigarette smoking. The risk for onset of chewing tobacco as a function of age was notably different from the risk for onset of cigarette smoking. This finding suggests that chewing tobacco prevention programs need to include men across the life span, not just adolescents as is typical with smoking prevention programs.

Consistent with previous research, our data indicated that rural status is a prime determinant of ever using chewing tobacco regularly.8,9 Compared with the overall rate (10.7%), urban men had a lower risk for regular use, especially urban men not living in the South. Among rural men, the most striking findings were the particularly high rate of regular lifetime use for lower-income Black and White men (33.3%) and the particularly low rate for lower-income Mexican American men (7.7%). Race/ethnicity did not distinguish among higher-income rural men who also had high rates of lifetime use (14.9%). These results suggest that if prevention resources are limited, they should be directed primarily to rural, lower-income Black and White men and secondarily to rural, higher-income men regardless of race/ethnicity.

Data on quitting indicated that nearly three quarters of the lifetime regular users not living in the South had successfully quit, whereas far fewer southern men had quit. More cessation programs may be available in nonsouthern regions, or there may be fewer social and cultural supports for continued use of chewing tobacco. These data suggest that lowering prevalence rates will require a more concerted effort in the South. In addition, southern men who began using chewing tobacco during later adulthood need special outreach efforts. Although the sample size for this group was small and argues for caution when interpreting the results, several hypotheses regarding the low quit rates among these men with late onset are tenable. Perhaps because of their late start, these men have not had time to quit. Alternatively, their late start and relatively low rate of cigarette smoking may mean that they had fewer reasons to think about quitting (e.g., few health problems related to tobacco use). Other research has suggested that chewers who are motivated to quit are more dependent on nicotine, have received advice to quit from health professionals, and have experienced more tobacco-related health problems compared with chewers who are less interested in quitting.20

Education, often a key predictor of tobacco use and other health risk factors, was not a significant determinant of risk groups in the signal detection analysis, but it was a distinguishing factor in the profiling of subgroups. The highest risk groups for regular use and not quitting both had very low educational attainment; two thirds never completed high school. These men also had lower incomes, were older, and were more likely to be from the rural South than were the men in the other subgroups. These findings have implications for intervention efforts. For example, prevention and cessation programs may need to pay particular attention to issues of literacy, community resources, and social and cultural norms.

Survival analyses indicated that the risk for onset of regular chewing tobacco use continued well into middle age and beyond, whereas the risk for onset of regular cigarette smoking became negligible by middle age. If chewing onset occurs across the life span, there are many implications for interventions. For example, prevention messages must address middle-aged and older men, in addition to adolescents and young adults. Because smoking prevention programs focus primarily on adolescents, chewing tobacco programs that are offered only in tandem with school- or community-based smoking programs will likely miss a substantial number of men who initiate use during adulthood. Chewing tobacco prevention and cessation programs are needed beyond high school, vocational school, and college and in settings outside of school. The workplace is a viable setting for delivering programs to middle-aged and older men. In addition, some research suggests that cessation interventions located at the site of health care services, such as dentist offices, can be successful with chewing tobacco users.21,22

Overall, more research establishing the efficacy of chewing tobacco cessation interventions of all kinds (e.g., group counseling, self-help, nicotine replacement therapies) is needed to augment the relatively small body of literature in this area.21–27

The results and conclusions of this analysis were subject to limitations of the NHANES III data. NHANES III was a cross-sectional survey and did not allow inferences about developmental or causal pathways. Rural status classification was imprecise because the disseminated database collapsed a 10-point rural and urban continuum code into 2 categories to prevent identification of the sampled counties. The interview was designed for breadth, not depth of coverage (e.g., no definition of “regular use of chewing tobacco” was included), and like other self-report interview methodologies, it was subject to recall bias (e.g., potential underreporting of current chewing tobacco use). This study analyzed data for White, Black, and Mexican American men only, which limits the generalizability of results, but NHANES III included large samples of adults in each racial/ethnic group from households at both the upper and the lower levels of educational attainment, thus strengthening the validity of findings related to socioeconomic status.

This study used signal detection and survival analysis methodologies to identify specific groups of men at risk for using chewing tobacco and not quitting and found evidence that the ages at onset for chewing tobacco and for smoking were substantially different. However, both the signal detection and the survival analyses should be considered exploratory rather than hypothesis testing. Future studies need to establish the replicability of the subgroups of chewing tobacco users identified in this study and the findings related to chewing tobacco onset.


This research was supported by Public Health Service Grant R01 CA64285 from the National Institutes of Health, National Cancer Institute, to Dr Howard-Pitney (Co-Principal Investigator) and an Established Investigatorship Award from the American Heart Association and US Public Health Service Grant 1-RO3-HL-57100 from the National Heart, Lung, and Blood Institute to Dr Winkleby.

The authors gratefully acknowledge David Ahn for data analysis; Helena Kraemer for her biostatistical expertise; Stephen P. Fortmann, Lisa Henriksen, Gregory J. Norman, and Nina Schleicher for their valuable comments on an earlier draft of the paper; and Alana Koehler for technical assistance.


B. Howard-Pitney and M. A. Winkleby both contributed to the conceptualization of the study and planned and interpreted the analyses. B. HowardPitney wrote the paper, with substantial contribution from M. A. Winkleby.

Peer Reviewed


1. Tucker L. Use of smokeless tobacco, cigarette smoking, and hypercholesterolemia. Am J Public Health. 1989;79:1048–1050. [PMC free article] [PubMed]
2. Bolinder G, Alfredsson L, Englund A, deFaire U. Smokeless tobacco use and increased cardiovascular mortality among Swedish construction workers. Am J Public Health. 1994;84:399–404. [PMC free article] [PubMed]
3. Connolly G, Winn D, Hecht S, Henningfield J, Walker B Jr, Hoffman D. The reemergence of smokeless tobacco. N Engl J Med. 1986;314:1020–1027. [PubMed]
4. Mattson ME, Winn DM. Smokeless tobacco: association with increased cancer risk. NCI Monogr. 1989;8:13–16. [PubMed]
5. Daly M, Pierson W. The use of smokeless tobacco among basic airmen. Am J Health Promot. 1990;4:266–269. [PubMed]
6. Gritz E, Ksir C, McCarthy W. Smokeless tobacco use in the United States: present and future trends. Ann Behav Med. 1985;7:24–27.
7. Marcus AC, Crane LA, Shopland DR, Lynn WR. Use of smokeless tobacco in the United States: recent estimates from the Current Population Survey. NCI Monogr. 1989;8:17–23. [PubMed]
8. Centers for Disease Control and Prevention. Use of smokeless tobacco among adults—United States, 1991. MMWR Morb Mortal Wkly Rep. 1993;42(14):263–266. [PubMed]
9. Centers for Disease Control and Prevention. Surveillance for selected tobacco-use behaviors—United States, 1988–1994. MMWR Morb Mortal Wkly Rep. 1994;43(SS-3):1–43.
10. Centers for Disease Control. The Surgeon General's 1989 Report on Reducing the Health Consequences of Smoking: 25 years of progress. MMWR Morb Mortal Wkly Rep. 1989;38(suppl 2):1–32. [PubMed]
11. Centers for Disease Control and Prevention. Preventing tobacco use among young people: a report of the surgeon general. Executive summary. MMWR Morb Mortal Wkly Rep. 1994;43(RR-4):1–10. [PubMed]
12. Ernster VL. Advertising and promotion of smokeless tobacco products. NCI Monogr. 1989;8:87–94. [PubMed]
13. Rodu B. An alternative approach to smoking control [editorial]. Am J Med Sci. 1994;308:32–34. [PubMed]
14. National Center for Health Statistics. Plan and operation of the Third National Health and Nutrition Examination Survey, 1988–94. Vital Health Stat 1. 1994;No. 32:1–407. [PubMed]
15. McDowell A, Engel A, Massey JT, Maurer K. Plan and operation of the Second National Health and Nutrition Examination Survey, 1976–1980. Vital Health Stat 1. 1981;No. 15:1–144. [PubMed]
16. National Center for Health Statistics. Plan and operation of the Health and Nutrition Examination Survey, United States, 1971–1973. Vital Health Stat 1. 1977;No. 10b. [PubMed]
17. House JS, Kessler RC, Herzog AR. Age, socioeconomic status, and health. Milbank Q. 1990;68:383–411. [PubMed]
18. Kraemer HC. Evaluating Medical Tests: Objective and Quantitative Guidelines. Newbury Park, Calif: Sage Publications; 1992.
19. Breiman L, Friedman JH, Olshen RA, Stone CJ. Classification and Regression Trees. Belmont, Calif: Wadsworth International Group; 1984.
20. Howard-Pitney B, Fortmann SP, Killen JD. Generalizability of findings from a chewing tobacco cessation clinical trial. Nicotine Tob Res. 2001;3:347–352. [PubMed]
21. Stevens VJ, Severson H, Lichtenstein E, Little SJ, Leben J. Making the most of a teachable moment: a smokeless-tobacco cessation intervention in the dental office. Am J Public Health. 1995;85:231–235. [PMC free article] [PubMed]
22. Severson HH, Andrews JA, Lichtenstein E, Gordon JS, Barckley MF. Using the hygiene visit to deliver a tobacco cessation program: results of a randomized clinical trial. J Am Dent Assoc. 1998;129:993–999. [PubMed]
23. Hatsukami K, Severson H. Oral spit tobacco: addiction, prevention and treatment. Nicotine Tob Res. 1999;1:21–44. [PubMed]
24. Hatsukami D, Jensen J, Allen S, Grillo M, Bliss R. Effects of behavioral and pharmacological treatment on smokeless tobacco users. J Consult Clin Psychol. 1996;64:153–161. [PubMed]
25. Hatsukami D, Anton D, Keenan R, Callies A. Smokeless tobacco abstinence effects and nicotine gum dose. Psychopharmacology (Berl).1992;106:60–66. [PubMed]
26. Howard-Pitney B, Killen JD, Fortmann SP. Quitting chew: results from a randomized trial using nicotine patches. Exp Clin Psychopharmacol. 1999;7:362–371. [PubMed]
27. Andrews JA, Severson HH, Lichtenstein E, Gordon JS, Barckley MF. Evaluation of a dental office tobacco cessation program: effects on smokeless tobacco use. Ann Behav Med. 1999;21:48–53. [PubMed]

Articles from American Journal of Public Health are provided here courtesy of American Public Health Association
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


  • MedGen
    Related information in MedGen
  • PubMed
    PubMed citations for these articles

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...