Nicotine & Tobacco Research, 2015, 719–726 doi:10.1093/ntr/ntu215 Original investigation Advance Access publication October 21, 2014
Original investigation
An Evaluation of Fruit and Vegetable Consumption and Cigarette Smoking Among Youth Downloaded from http://ntr.oxfordjournals.org/ at University of Otago on November 14, 2015
Jeffrey P. Haibach PhD, MPH1, Gregory G. Homish PhD1, R. Lorraine Collins PhD1, Christine B. Ambrosone PhD2, Gary A. Giovino PhD, MS1 Department of Community Health and Health Behavior, School of Public Health and Health Professions, University at Buffalo, The State University of New York, Buffalo, NY; 2Department of Cancer Prevention and Control, Roswell Park Cancer Institute, Buffalo, NY 1
Corresponding Author: Jeffrey P. Haibach, PhD, MPH, Department of Community Health and Health Behavior, School of Public Health and Health Professions, University at Buffalo, The State University of New York, 311 Kimball Tower, Buffalo, NY 14214-8028, USA. Telephone: 716-829-6100; Fax: 716-829-6040; E-mail: [email protected]
Abstract Introduction: Research across the past 4 decades has supported a cross-sectional association between adult cigarette smoking and lower fruit and vegetable consumption (FVC), and emerging research suggests higher FVC may predict cessation. Among youth, findings are limited to a few cross-sectional studies with somewhat mixed results. Here we evaluated the FVC-smoking association among youth both cross-sectionally and longitudinally. Methods: We analyzed data from a subsample of the National Longitudinal Survey of Youth 1979: Child and Young Adult. The subsample included adolescents aged 14–18 years at baseline in the year 2004. Multivariable cross-sectional analyses assessed whether baseline FVC was associated with smoking frequency among ever-smokers (n = 578). Longitudinally, the study assessed whether baseline FVC predicted smoking progression among baseline never-smokers who tried a cigarette by 4-year follow-up (n = 388). Multivariable regression models adjusted for age, gender, race/ethnicity, parental education, and health behavior orientation. Results: Cross-sectionally, youth who consumed fruit ≥2 times per day were 53% less likely (RR = 0.47; p < .05) than those who typically did not consume fruit to be in a higher smoking frequency category. Longitudinally, the fruit consumption and smoking association was not significant (RR = 0.61; p = .282). There were no significant associations observed between vegetable consumption and smoking. Conclusions: Fruit consumption, but not vegetable consumption, was inversely associated with smoking frequency cross-sectionally but not longitudinally. Further research is needed to provide information on the consistency of the FVC-smoking relationship among youth and may help to elucidate possible explanatory mechanisms.
Introduction Nine out of 10 cigarette smokers in the United States try their first cigarette by age 18, and at least one-third of them will die early from smoking attributable disease.1 Research further suggests that
changes in young adult smoking prevalence are similar to adolescent smoking prevalence, but found to lag two to three years in a cohort effect.2 Thus, reducing adolescent smoking uptake would be expected to reduce adult smoking prevalence and premature mortality attributable to smoking.
© The Author 2014. Published by Oxford University Press on behalf of the Society for Research on Nicotine and Tobacco. All rights reserved. For permissions, please e-mail: [email protected].
719
Nicotine & Tobacco Research, 2015, Vol. 17, No. 6
720
Methods Participants This study was performed using secondary data analysis of a national population-based dataset, the National Longitudinal Survey of Youth 1979: Child and Young Adult (NLSY79-CYA). The NLSY79CYA consists of children born to females of an original cohort study that began in 1979 of U.S. participants aged 14–22 years, the National Longitudinal Survey of Youth.22 There was an oversample of Hispanic and Black persons in the original parent cohort. Data for the NLSY79-CYA were collected through self-report interview surveys completed in person or by telephone with a trained interviewer using computer-assisted interviewing techniques. Surveys were conducted biennially with completion rates ranging from 83.0% to 99.9%.22 The two subsamples for this study were limited to 14–18 year olds at baseline. The cross-sectional sample included those who had ever tried a cigarette at baseline in the year 2004 for smoking frequency analyses (n = 578; 56.7% male; Mage = 16.7 years, SD = 1.3; see Table 1 for demographic and health behavior descriptive statistics). The subsample of ever-smokers was selected because a true biological association between FVC and smoking would be exhibited by resistance to uptake on exposure to nicotine and other addictive properties of cigarettes during experimentation, not necessarily whether a youth tried a cigarette or not. Including never-smokers would be assessing, at least in part, whether someone tried a cigarette or not, which would more likely be explained by non-biological environmental etiology shared by both FVC and smoking such as SES. The longitudinal sample consisted of never-smokers at baseline (58.5% male; baseline Mage = 16.0 years, SD = 2.1) who ever tried a cigarette by 4-year follow-up for level of smoking progression (year 2008; n = 388). The never-smoker subsample was chosen for longitudinal analyses to assess the direction of the association with FVC assessed prior to ever smoking to remove the possibility of prior smoking predicting baseline FVC. The baseline age group of 14–18 year olds was selected as FVC data were not collected in the younger age groups surveyed, and the majority of smokers have tried their first cigarette by the age of 18 years.1
Measures Fruit and Vegetable Independent Variables Independent variables of interest included frequency of consumption of (a) fruits, (b) vegetables, and (c) an aggregate fruit and vegetable variable. The NLSY79-CYA used two questions to assess FVC in 2004, one for fruit and one for vegetables. They were: (a) “In a typical week, how many times do you eat fruit?” and (b) “In a typical week, how many times do you eat vegetables other than French fries or potatoes?” The seven response categories were: (a) “I do not typically eat (fruit/vegetables),” (b) “1–3 times per week,” (c) “4–6 times per week,” and categories 4 through 7 represent 1, 2, 3, and ≥4 “times per day,” respectively. Brief questions assessing FVC in times “per week” or “per day” are widely used and validated to estimate the level of FVC in population surveys.23–25 Responses were converted into four categories that were a balance between the nearest quartile and conceptual cut-point that the data would allow, consistent with nutritional epidemiology methods.26 This also allows for more sensitive analyses to examine a threshold or dose-response effect, whereas prior studies1,13,14 assessing the FVC-smoking association among youth used factor analysis, comparison of means, or dichotomous FVC cut-points. Individual fruit and vegetable
Downloaded from http://ntr.oxfordjournals.org/ at University of Otago on November 14, 2015
While regulatory, pharmaceutical, and economic efforts have been effective at reducing smoking prevalence, additional methods or factors that target biological processes may increase the rate of decline in smoking uptake and ultimately reduce long term prevalence. One possible factor that has consistently been found to be associated with higher smoking among adults, has direct biological implications for smoking, and has yet to be reported as relevant for smoking prevention is low fruit and vegetable consumption (FVC).3–12 Existing research on the association between FVC and smoking among adolescents is limited, with somewhat mixed results. In a study of smoking and weight loss attempts among 1,132 adolescents participating in the third cycle of the National Health and Nutrition Examination Survey (NHANES III), researchers found that adolescent smokers had lower FVC compared to nonsmokers.13 In a British study of 932 students aged 12–15 years, smokers scored lower on a composite index of certain healthier foods, largely fruits and vegetables.14 Analysis of data from the 1999–2009 Youth Risk Behavior Surveillance System (YRBSS) did not find a significant cross-sectional FVC-smoking association among high school seniors.1 Both the YRBSS and NHANES studies were focused on the association between weight control and smoking. In these studies, the FVCsmoking association was framed in terms of weight control but not discussed in regard to a direct FVC-smoking association. The Coulson study described possible health behavior covariance among diet, smoking, and exercise as likely being explained by common behavioral factors of general health motivation and differing values placed on health between individuals. None of the three cross-sectional studies address the direction of the FVC-smoking association, nor is directionality hypothesized in the articles. Prior studies among adults have suggested that the association is explained by a common etiology of socioeconomic status (SES) or health consciousness. However, the Strauss NHANES III study adjusted for family income and at least a few of the adult studies have also adjusted for SES indicators.9,10,15 One longitudinal study of adult cessation adjusted for SES indicators of education and household income as well as general health behavior orientation,16 and the FVC-smoking association persisted. If low FVC is a risk factor for smoking, this represents a substantial public health risk as the majority of adolescents in the United States have low FVC. In 2011, 4.8% of high school students did not consume fruit or fruit juice within the prior week and only 34.0% consumed them at least two times per day.17 Also in 2011, only 15.3% ate vegetables at least three times per day and 5.7% did not eat vegetables at all. In evaluating FVC by cup per day recommendations18 based on calorie requirements for specific age, gender, and physical activity levels, full FVC recommendations were met by only 0.9% of adolescents aged 12–18 years and 2.2% of adults aged ≥19 years.19 Food choices such as FVC also generally remain consistent from childhood through adulthood.20,21 Low FVC among youth may be an additional risk factor for consideration in smoking prevention. Given the limited research with mixed findings on the FVCsmoking association among youth, this study first tested whether higher FVC among adolescents predicted a lower smoking frequency in days per week among those who ever tried a cigarette. This adds a national population-based study to the limited and mixed cross-sectional research on FVC and smoking among adolescents. To explore directionality, we then tested whether baseline FVC predicted 4-year follow-up smoking frequency among those who had never smoked a cigarette at baseline.
Nicotine & Tobacco Research, 2015, Vol. 17, No. 6
721
Table 1. Sample Characteristics Cohort sample (never-smokers at baseline)
Ever-smoked, n = 578
Smoked by 4-year follow-up, n = 388
1.2 (1.9) 6.5 (6.9) 6.3 (6.3) 12.8 (11.2) 19.9 7.3 8.0
1.6 (2.1) 7.9 (7.6) 6.9 (6.6) 14.7 (12.4) 28.2 7.0 8.5
16.7 (1.3)
16.0 (1.4)
56.7
58.5
25.8 20.8 53.5 74.0 41.3 18.2 60.9
25.3 31.2 43.6 80.4 47.9 3.1 13.4
DPW = days per week; MPD = minutes per day; TPD = times per day; TPW = times per week. a Smoking frequency/progression was modeled as an ordinal distribution with past 30 day smoking categories of: (a) no smoking, (b) smoking less than weekly, (c) 1–2 days per week (DPW), (d) 3–4 DPW, (e) 5–6 DPW, and (f) daily smoking. b Illicit drug use includes ever using amphetamines, cocaine, crack, downers, hallucinogens, heroin, marijuana, MDMA (ecstasy), sniffing glue, steroids, stimulants, or any other drugs not “prescribed” to the respondent (not including alcohol or tobacco).
variables were categorized as: (a) no typical weekly consumption, (b) 1–3 times per week, (c) 1 time per day (includes 4–6 times per week rounded to 1 time per day to err closer to a quartile distribution), and (d) ≥2 times per day. Aggregate FVC was categorized as: (a) less than daily consumption, (b) 1 time per day, (c) 2–3 times per day, and (d) ≥4 times per day. Additional analyses were performed with dichotomous FVC variables at cut-points most aligned with dietary recommendations and Centers for Disease Control and Prevention assessment (less than or ≥2 times per day for fruit, ≥3 times per day for vegetables, and ≥5 times per day for aggregate FVC).17 Smoking Dependent Variables Smoking frequency was assessed in the NLSY79-CYA by asking those screened as ever-smokers, “During the last 30 days, how often, if ever, have you smoked cigarettes on average?” The six response categories were: (a) “never in the last 30 days,” (b) “less than once a week,” (c) “1 or 2 days per week,” (d) “3 or 4 days per week,” (e) “5 or 6 days per week,” and (f) “every day.” Using frequency of smoking in the last 30 days is a relatively standard option in population research27 and smoking self-report in population based surveys has been found to provide smoking estimates comparable to those based on clinical measures of serum biomarkers.28–30 Smoking frequency was analyzed in its original surveyed categories and considered a measure of smoking progression in the longitudinal analyses. Adjustment Variables Demographic variables adjusted for in statistical models included age (continuous), gender (male, female), race/ethnicity (Hispanic, Black, and non-Hispanic/non-Black as per NLSY79-CYA data reporting),
and parental education as an SES indicator (
Original investigation
An Evaluation of Fruit and Vegetable Consumption and Cigarette Smoking Among Youth Downloaded from http://ntr.oxfordjournals.org/ at University of Otago on November 14, 2015
Jeffrey P. Haibach PhD, MPH1, Gregory G. Homish PhD1, R. Lorraine Collins PhD1, Christine B. Ambrosone PhD2, Gary A. Giovino PhD, MS1 Department of Community Health and Health Behavior, School of Public Health and Health Professions, University at Buffalo, The State University of New York, Buffalo, NY; 2Department of Cancer Prevention and Control, Roswell Park Cancer Institute, Buffalo, NY 1
Corresponding Author: Jeffrey P. Haibach, PhD, MPH, Department of Community Health and Health Behavior, School of Public Health and Health Professions, University at Buffalo, The State University of New York, 311 Kimball Tower, Buffalo, NY 14214-8028, USA. Telephone: 716-829-6100; Fax: 716-829-6040; E-mail: [email protected]
Abstract Introduction: Research across the past 4 decades has supported a cross-sectional association between adult cigarette smoking and lower fruit and vegetable consumption (FVC), and emerging research suggests higher FVC may predict cessation. Among youth, findings are limited to a few cross-sectional studies with somewhat mixed results. Here we evaluated the FVC-smoking association among youth both cross-sectionally and longitudinally. Methods: We analyzed data from a subsample of the National Longitudinal Survey of Youth 1979: Child and Young Adult. The subsample included adolescents aged 14–18 years at baseline in the year 2004. Multivariable cross-sectional analyses assessed whether baseline FVC was associated with smoking frequency among ever-smokers (n = 578). Longitudinally, the study assessed whether baseline FVC predicted smoking progression among baseline never-smokers who tried a cigarette by 4-year follow-up (n = 388). Multivariable regression models adjusted for age, gender, race/ethnicity, parental education, and health behavior orientation. Results: Cross-sectionally, youth who consumed fruit ≥2 times per day were 53% less likely (RR = 0.47; p < .05) than those who typically did not consume fruit to be in a higher smoking frequency category. Longitudinally, the fruit consumption and smoking association was not significant (RR = 0.61; p = .282). There were no significant associations observed between vegetable consumption and smoking. Conclusions: Fruit consumption, but not vegetable consumption, was inversely associated with smoking frequency cross-sectionally but not longitudinally. Further research is needed to provide information on the consistency of the FVC-smoking relationship among youth and may help to elucidate possible explanatory mechanisms.
Introduction Nine out of 10 cigarette smokers in the United States try their first cigarette by age 18, and at least one-third of them will die early from smoking attributable disease.1 Research further suggests that
changes in young adult smoking prevalence are similar to adolescent smoking prevalence, but found to lag two to three years in a cohort effect.2 Thus, reducing adolescent smoking uptake would be expected to reduce adult smoking prevalence and premature mortality attributable to smoking.
© The Author 2014. Published by Oxford University Press on behalf of the Society for Research on Nicotine and Tobacco. All rights reserved. For permissions, please e-mail: [email protected].
719
Nicotine & Tobacco Research, 2015, Vol. 17, No. 6
720
Methods Participants This study was performed using secondary data analysis of a national population-based dataset, the National Longitudinal Survey of Youth 1979: Child and Young Adult (NLSY79-CYA). The NLSY79CYA consists of children born to females of an original cohort study that began in 1979 of U.S. participants aged 14–22 years, the National Longitudinal Survey of Youth.22 There was an oversample of Hispanic and Black persons in the original parent cohort. Data for the NLSY79-CYA were collected through self-report interview surveys completed in person or by telephone with a trained interviewer using computer-assisted interviewing techniques. Surveys were conducted biennially with completion rates ranging from 83.0% to 99.9%.22 The two subsamples for this study were limited to 14–18 year olds at baseline. The cross-sectional sample included those who had ever tried a cigarette at baseline in the year 2004 for smoking frequency analyses (n = 578; 56.7% male; Mage = 16.7 years, SD = 1.3; see Table 1 for demographic and health behavior descriptive statistics). The subsample of ever-smokers was selected because a true biological association between FVC and smoking would be exhibited by resistance to uptake on exposure to nicotine and other addictive properties of cigarettes during experimentation, not necessarily whether a youth tried a cigarette or not. Including never-smokers would be assessing, at least in part, whether someone tried a cigarette or not, which would more likely be explained by non-biological environmental etiology shared by both FVC and smoking such as SES. The longitudinal sample consisted of never-smokers at baseline (58.5% male; baseline Mage = 16.0 years, SD = 2.1) who ever tried a cigarette by 4-year follow-up for level of smoking progression (year 2008; n = 388). The never-smoker subsample was chosen for longitudinal analyses to assess the direction of the association with FVC assessed prior to ever smoking to remove the possibility of prior smoking predicting baseline FVC. The baseline age group of 14–18 year olds was selected as FVC data were not collected in the younger age groups surveyed, and the majority of smokers have tried their first cigarette by the age of 18 years.1
Measures Fruit and Vegetable Independent Variables Independent variables of interest included frequency of consumption of (a) fruits, (b) vegetables, and (c) an aggregate fruit and vegetable variable. The NLSY79-CYA used two questions to assess FVC in 2004, one for fruit and one for vegetables. They were: (a) “In a typical week, how many times do you eat fruit?” and (b) “In a typical week, how many times do you eat vegetables other than French fries or potatoes?” The seven response categories were: (a) “I do not typically eat (fruit/vegetables),” (b) “1–3 times per week,” (c) “4–6 times per week,” and categories 4 through 7 represent 1, 2, 3, and ≥4 “times per day,” respectively. Brief questions assessing FVC in times “per week” or “per day” are widely used and validated to estimate the level of FVC in population surveys.23–25 Responses were converted into four categories that were a balance between the nearest quartile and conceptual cut-point that the data would allow, consistent with nutritional epidemiology methods.26 This also allows for more sensitive analyses to examine a threshold or dose-response effect, whereas prior studies1,13,14 assessing the FVC-smoking association among youth used factor analysis, comparison of means, or dichotomous FVC cut-points. Individual fruit and vegetable
Downloaded from http://ntr.oxfordjournals.org/ at University of Otago on November 14, 2015
While regulatory, pharmaceutical, and economic efforts have been effective at reducing smoking prevalence, additional methods or factors that target biological processes may increase the rate of decline in smoking uptake and ultimately reduce long term prevalence. One possible factor that has consistently been found to be associated with higher smoking among adults, has direct biological implications for smoking, and has yet to be reported as relevant for smoking prevention is low fruit and vegetable consumption (FVC).3–12 Existing research on the association between FVC and smoking among adolescents is limited, with somewhat mixed results. In a study of smoking and weight loss attempts among 1,132 adolescents participating in the third cycle of the National Health and Nutrition Examination Survey (NHANES III), researchers found that adolescent smokers had lower FVC compared to nonsmokers.13 In a British study of 932 students aged 12–15 years, smokers scored lower on a composite index of certain healthier foods, largely fruits and vegetables.14 Analysis of data from the 1999–2009 Youth Risk Behavior Surveillance System (YRBSS) did not find a significant cross-sectional FVC-smoking association among high school seniors.1 Both the YRBSS and NHANES studies were focused on the association between weight control and smoking. In these studies, the FVCsmoking association was framed in terms of weight control but not discussed in regard to a direct FVC-smoking association. The Coulson study described possible health behavior covariance among diet, smoking, and exercise as likely being explained by common behavioral factors of general health motivation and differing values placed on health between individuals. None of the three cross-sectional studies address the direction of the FVC-smoking association, nor is directionality hypothesized in the articles. Prior studies among adults have suggested that the association is explained by a common etiology of socioeconomic status (SES) or health consciousness. However, the Strauss NHANES III study adjusted for family income and at least a few of the adult studies have also adjusted for SES indicators.9,10,15 One longitudinal study of adult cessation adjusted for SES indicators of education and household income as well as general health behavior orientation,16 and the FVC-smoking association persisted. If low FVC is a risk factor for smoking, this represents a substantial public health risk as the majority of adolescents in the United States have low FVC. In 2011, 4.8% of high school students did not consume fruit or fruit juice within the prior week and only 34.0% consumed them at least two times per day.17 Also in 2011, only 15.3% ate vegetables at least three times per day and 5.7% did not eat vegetables at all. In evaluating FVC by cup per day recommendations18 based on calorie requirements for specific age, gender, and physical activity levels, full FVC recommendations were met by only 0.9% of adolescents aged 12–18 years and 2.2% of adults aged ≥19 years.19 Food choices such as FVC also generally remain consistent from childhood through adulthood.20,21 Low FVC among youth may be an additional risk factor for consideration in smoking prevention. Given the limited research with mixed findings on the FVCsmoking association among youth, this study first tested whether higher FVC among adolescents predicted a lower smoking frequency in days per week among those who ever tried a cigarette. This adds a national population-based study to the limited and mixed cross-sectional research on FVC and smoking among adolescents. To explore directionality, we then tested whether baseline FVC predicted 4-year follow-up smoking frequency among those who had never smoked a cigarette at baseline.
Nicotine & Tobacco Research, 2015, Vol. 17, No. 6
721
Table 1. Sample Characteristics Cohort sample (never-smokers at baseline)
Ever-smoked, n = 578
Smoked by 4-year follow-up, n = 388
1.2 (1.9) 6.5 (6.9) 6.3 (6.3) 12.8 (11.2) 19.9 7.3 8.0
1.6 (2.1) 7.9 (7.6) 6.9 (6.6) 14.7 (12.4) 28.2 7.0 8.5
16.7 (1.3)
16.0 (1.4)
56.7
58.5
25.8 20.8 53.5 74.0 41.3 18.2 60.9
25.3 31.2 43.6 80.4 47.9 3.1 13.4
DPW = days per week; MPD = minutes per day; TPD = times per day; TPW = times per week. a Smoking frequency/progression was modeled as an ordinal distribution with past 30 day smoking categories of: (a) no smoking, (b) smoking less than weekly, (c) 1–2 days per week (DPW), (d) 3–4 DPW, (e) 5–6 DPW, and (f) daily smoking. b Illicit drug use includes ever using amphetamines, cocaine, crack, downers, hallucinogens, heroin, marijuana, MDMA (ecstasy), sniffing glue, steroids, stimulants, or any other drugs not “prescribed” to the respondent (not including alcohol or tobacco).
variables were categorized as: (a) no typical weekly consumption, (b) 1–3 times per week, (c) 1 time per day (includes 4–6 times per week rounded to 1 time per day to err closer to a quartile distribution), and (d) ≥2 times per day. Aggregate FVC was categorized as: (a) less than daily consumption, (b) 1 time per day, (c) 2–3 times per day, and (d) ≥4 times per day. Additional analyses were performed with dichotomous FVC variables at cut-points most aligned with dietary recommendations and Centers for Disease Control and Prevention assessment (less than or ≥2 times per day for fruit, ≥3 times per day for vegetables, and ≥5 times per day for aggregate FVC).17 Smoking Dependent Variables Smoking frequency was assessed in the NLSY79-CYA by asking those screened as ever-smokers, “During the last 30 days, how often, if ever, have you smoked cigarettes on average?” The six response categories were: (a) “never in the last 30 days,” (b) “less than once a week,” (c) “1 or 2 days per week,” (d) “3 or 4 days per week,” (e) “5 or 6 days per week,” and (f) “every day.” Using frequency of smoking in the last 30 days is a relatively standard option in population research27 and smoking self-report in population based surveys has been found to provide smoking estimates comparable to those based on clinical measures of serum biomarkers.28–30 Smoking frequency was analyzed in its original surveyed categories and considered a measure of smoking progression in the longitudinal analyses. Adjustment Variables Demographic variables adjusted for in statistical models included age (continuous), gender (male, female), race/ethnicity (Hispanic, Black, and non-Hispanic/non-Black as per NLSY79-CYA data reporting),
and parental education as an SES indicator (
