4 lancet

7/28/2019 4 Lancet

1/6

7/28/2019 4 Lancet

2/6

Articles

590 www.thelancet.com Vol 367 February 18, 2006

and eighth-grade students in the 32 schools. Wepostulated that students who were in government

schools, boys, older in age, and in the eighth grade,would be using more tobacco than those who were inprivate schools, girls, younger in age, and in the sixthgrade.

All students in the sixth and eighth grades (n=12 484)in the 32 schools were eligible and were invited toparticipate in the baseline survey in the summer of 2004.Informed passive parental consent and active studentassent procedures were used. Letters were sent homefrom the schools to parents of all eligible students.Parents were asked to return a card if they did not wanttheir child to participate in the survey. Students were reada form at the time of the survey ensuring confidentialityand also that their standing with their school and with theproject would not be jeopardised by not participating.

Every student signed an assent form if they agreed toparticipate. The institutional review board at theUniversity of Minnesota (Minneapolis, MN, USA) andthe Indian Independent Ethics Committee (Mumbai,India) approved the study protocol.

ProceduresThe tobacco use survey was a self-administered paperand pencil survey administered in school classrooms bytrained survey interviewers from the project. Thequestions about tobacco use were adapted from theGlobal Youth Tobacco Survey, which has been used ineighth-grade to tenth-grade students.11,12 To ensureappropriate adaptation for sixth-grade students and ourpopulation, we did 48 focus groups with 435 students togather information on their understanding of tobacco,tobacco use, and psychosocial predictive factors.13 Wethen developed the questionnaire on the basis of thisinformation, the Global Youth Tobacco Survey, andprevious surveys. The draft questionnaire was translatedfrom English and back-translated into Hindi and Tamil.All private schools in both cities had English versions.The government schools versions were in Hindi inDelhi and in Tamil in Chennai. We administered thedraft questionnaire to small groups of sixth-grade andeighth-grade students (n=60) in English, Hindi, andTamil in government and private schools in Delhi andChennai, and then discussed each question with them,

and modified the questionnaire accordingly. Finally, wepiloted the survey in English and Hindi with235 students in private and government schools in Delhiin order to pilot survey implementation (includingstudent questions and concerns during administration)and assess the psychometric properties of the survey.

Tobacco use items measured ever use and current useof chewing tobacco, cigarettes, and bidis (hand-rolledcigarettes). Current use of tobacco was measured by thequestions: During the last 30 days, did you (chewtobacco in any form?) (smoke one or more bidis?)(smoke one or more cigarettes?). The responsecategories were yes or no. Ever use of tobacco wasmeasured by the questions: How old were you when

you first (chewed tobacco in any form?) (put a litcigarette in your mouth?) (put a lit bidi in your mouth?).The response categories were I have never (chewedtobacco) (put a lit cigarette in my mouth) (put a lit bidi inmy mouth), or a specific age ranging from 7 years orless to 16 years or more. These response categories werecollapsed to create a dichotomous variable: no use versusever use of tobacco.

Additionally, psychosocial factors that are associatedwith tobacco use among young people in the USA wereassessed with scales that measured intentions to usetobacco in the future, social susceptibility to use tobacco,reasons to use tobacco, and normative expectationsconcerning tobacco use.6,14-17 All scales were created byadding up the scores of the responses to individual items.

The two intentions scales each included four items: Doyou think you will try chewing tobacco (smokingcigarettes or bidis) in the next month? In the next year?When you enter college? When you are an adult?. Eachitem had four response categories including: surely yes(3), maybe yes (2), maybe no (1), and surely no (0). Thescale range was 012 and the coefficients for the inten-tions scales were 085 (chewing) and 087 (smoking).The two social susceptibility scales each included fouritems: If one of your close friends gave you chewingtobacco (a cigarette or bidi), would you chew (smoke) it?If a group of friends gave you . . . ? If one of your familymembers gave you . . . ? If someone at a party gaveyou . . . ? Each item had four response categories:surely yes (3), maybe yes (2), maybe no (1), and surely no(0). The scale range was 012 and the coefficients forthe two social susceptibility scales were 087 (chewing)and 088 (smoking). The reasons to use tobacco scale wasmeasured by six items that addressed whether chewingor smoking was fashionable, fun to do with friends,grown up and brave, a way to reduce boredom, attractiveto friends who are boys, and attractive to friends who aregirls. Each item had four response categories: surely yes(3), maybe yes (2), maybe no (1), and surely no (0). Thescale range was 018 and the coefficient for the reasonsto use tobacco scale was 073. The final scale addressedsocial norms (normative expectations) concerningtobacco use and was measured by six items: If you were

32 schools

16 Delhi

8 government 8 private

2 girls

2 boys

4 co-educational

8 co-educational

16 Chennai

8 government 8 private

2 girls

2 boys

4 co-educational

8 co-educational

Figure: Project MYTRI study design

7/28/2019 4 Lancet

3/6

Articles

to use tobacco, do you think . . . Your close friends wouldlike it? Your parents would like it? Your teachers wouldlike it? Your relatives/neighbours would like it? Boys inyour school/neighbourhood would like it? Girls in yourschool/neighbourhood would like it? Each item had fourresponse categories: surely yes (0), maybe yes (1), maybeno (2), and surely no (3). The scale range was 018 andthe coefficient for the normative expectations scale was091. To assess the validity of the psychosocial scales, therelations between the scales and tobacco use measureswere assessed among all students using a series ofregression models. All the scales were significantlyassociated with ever use of tobacco (p005).

Trained survey interviewers from the Project MYTRI

staff introduced the questionnaire in the classrooms,then allowed students to complete it at their desk with apencil that we provided. Survey interviewers answeredany questions as they arose during the surveyadministration, including queries related to theclarification of questions on the survey instrument.Students were given unique identification codes to assureconfidentiality. Teachers remained in the classrooms butdid not participate in the survey administration.

Statistical analysisA series of mixed-effects regression models were used toassess differences in rates of tobacco use and psycho-social scales by relevant demographic factors, includingcity (Delhi vs Chennai), type of school (private vs

government), sex (boys vs girls), grade (sixth vs eighthgrade), and age (11 years vs 12 years vs 13 years vs14 years). This kind of regression model is the mostappropriate, in view of the nested study design, as itaccounts for variability between both students andschools.18 All comparisons between grade levels wereadjusted for, when not stratified by, other demographicfactors (city, school type, sex, age). In examininginteractions between grade and other demographicfactors, the only significant interaction was between ageand grade, so comparisons between grade levels wereadditionally adjusted for the age*grade interaction. Allanalyses were done with SAS (version 8.80) statisticalsoftware.

Role of the funding sourceThe sponsor of the study had no role in study design,data collection, data analysis, data interpretation, orwriting of the report. The corresponding author had fullaccess to all the data in the study and had finalresponsibility for the decision to submit for publication.

Results94% of the sample participated (n=11 748). Of theremainder, 44% were absent on the initial and make-upsurvey days (we came in on a second day to every schoolto survey those who were absent on the initial surveyday); 15% did not participate because of parent or

student refusal. After exclusion of inconsistentresponders (1%), that is, those with four or moreinconsistent responses on the survey (such as reportingthat they had used tobacco in the past month, but hadnot ever used in their lives), the analysis sample size was11 642. The sample analysed included 5889 (506%)from Delhi, 4489 (386%) from private schools, 6386(549%) who were male, and 6165 (529%) who were inthe sixth grade. Mean age was 1121 years (range 1016years) for students in the sixth grade and 1292 years(1016 years) for those in the eighth grade.

Overall, 1667 (147%) students had ever used tobacco;1242 (108%) had chewed tobacco, 851 (74%) hadsmoked cigarettes, and 796 (7%) had smoked bidis.

520 (46%) were current tobacco users, with 346 (3%)currently using chewing tobacco, 163 (14%) cigarettes,and 180 (16%) bidis. Significant differences in ever useof any kind of tobacco were found by type of school, sex,age, and grade level. 1237 (173%) students at govern-ment schools and 476 (106%) at private schools hadever used tobacco. 1086 (17%) boys and 568 (108%)girls had used tobacco. Of students aged 10 years oryounger, 96 (6%) had used tobacco, compared with282 (93%) aged 11 years, 416 (149%) aged 12 years,454 (176%) aged 13 years, and 346 (219%) aged14 years or older. 1529 (248%) sixth-grade students hadever used tobacco, compared with 509 (93%) eighth-grade students.

These differences by demographic factors were in the

expected direction except for the difference betweensixth-grade and eighth-grade students. Students in thesixth grade were significantly more likely to use allforms of tobacco than students in the eighth grade(table 1), in each city, in government schools, and forboth sexes (table 2). Sixth-grade students tobacco usewas, overall, two to four times that of eighth-gradestudents. Students in the sixth grade were overall usingmore tobacco than those in the eighth grade in every agegroup. The increase in tobacco use by age was greater insixth-grade students than eighth-grade students(p=00003).

www.thelancet.com Vol 367 February 18, 2006 591

Sixth grade Eighth grade Ratio* p

(n=6165) (n=5477)

Ever use of tobacco

Chewing tobacco 190% (174206) 68% (5383) 28:1 00001

Smoking bidis 119% (106132) 34% (2246) 35:1 00001

Smoking cigarettes 125% (111139) 47% (3559) 27:1 00001

Any kind of tobacco 248% (230266) 93% (77109) 27:1 00001

Current use of tobacco

Chewing tobacco 45% (3753) 16% (0923) 28:1 00001

Smoking bidis 20% (1426) 09% (0414) 22:1 00009

Smoking cigarettes 20% (1426) 09% (0414) 22:1 00016

Any kind of tobacco 67% (5678) 29% (1939) 23:1 00001

Data in parentheses are 95% CI. Estimates generated from mixed-effects models adjusted for city, school type, sex, age, and

grade*age. *Compares prevalence of tobacco use in sixth grade with that in eighth grade.

Table 1: Differences in prevalence of tobacco use between sixth-grade and eighth-grade students

(n=11 642)

7/28/2019 4 Lancet

4/6

Articles


The only difference between grade levels that was notsignificant was for private schools. In further examiningthe interaction between age and grade by school type, thedifferences between sixth and eighth grades weresignificant for all ages in government schools, but onlyfor 12-year-olds in private schools (table 3). However, inthe private schools, only 33 students (14%) in the sixthgrade were aged 13 years or older, and only 55 students

(24%) in the eighth grade were aged 11 years oryounger. In view of these small numbers, thecomparisons by age for private schools were limited inmeaning. Notably, in 12-year-olds at private schools,sixth-grade students tobacco use was nearly four timesthat of eighth-grade students.

The significant age*grade interaction in governmentschools seemed to be driven by the clear increase intobacco use by age in sixth-grade students; this trend was

less apparent in eighth-grade students. In private schools,the significant interaction seemed to be driven bydifferent relations between grade and tobacco use by age.In 12-year-olds, sixth-grade students used more tobaccothan eighth-grade students, but in those aged 13 years or14 years and older, the prevalence was actually higher ineighth-grade students (although not significantly so).

Analyses of psychosocial factors also showedconsistently significant differences between grade levels(table 4). Students in the sixth grade also had greaterintentions to use tobacco in the future, more socialsusceptibility to use, more positive reasons to usetobacco, and more positive normative expectationsconcerning tobacco use than those in the eighth grade.These psychosocial risk factors were all associated with

significantly greater tobacco use in all students (p005,data not shown).

DiscussionIncreased grade level in school has been thought to bethe factor that most reliably predicts tobacco use amongyoung people,6 but we found the opposite in our datafrom Delhi and Chennai. Even though increased age wasalso associated with tobacco use, this trend was foundwithin rather than across grade levels. In fact, increasein tobacco use by age was significantly greater in thesixth-grade than in the eighth-grade cohort. Thisdifference was noted primarily in government schools,among students with lower socioeconomic status. Thesefindings are highly unusual and suggest that this groupof teenagers in urban India are just beginning to usetobacco at increased rates. Of particular concern is thevery high prevalence of ever having used tobacco(32%) in sixth-grade students aged 13 years and olderin government schools. Since early use of tobaccopredicts greater likelihood of addiction, longer lifetimeuse, and higher rates of lung cancer, these findings areof importance to public health.5,6

The results of the analyses of psychosocial factors arealso consistent with greater use among sixth-grade thaneighth-grade students, since these factors indicate thatsixth-grade students have a significantly greater riskprofile associated with onset of tobacco use in this age


Number Prevalence Number Prevalence

Private schools

Total 2326 2143

Age11 years 2090 110% (84 to 136) 52 110% (23 to 197) 10:1 09939

Age 12 years 203 215% (172 to 258) 934 59% (38 to 80) 36:1 00001

Age 13 years 22 30% (89 to 149) 1018 76% (56 to 96) 04:1 04513

Age14 years 11 38% (169 to 245) 139 70% (20 to 160) 05:1 07414

Government schools

Total 3811 3306

Age11 years 2289 215% (182 to 248) 202 121% (60 to 182) 18:1 00008

Age 12 years 814 258% (231 to 285) 841 61% (33 to 89) 42:1 00001

Age 13 years 426 325% (288 to 362) 1114 98% (74 to 122) 33:1 00001

Age14 years 282 362% (312 to 412) 1149 157% (133 to 181) 23:1 00001

Data in parentheses are 95% CI. Estimates generated from mixed-effects models adjusted for city and sex. 56 students did not

provide their age, so the analysis sample was 11 586. *Ratio compares prevalence of tobacco use in sixth grade with that in

eighth grade.

Table 3: Differences in prevalence of ever use of any kind of tobacco, by grade, age, and type of school

(n=11 642)

Sixth grade Eighth grade p

(n=6165) (n=5477)

Intentions to chew 155 (550) 076 (444) 00001

Intentions to smoke 097 (393) 045 (296) 00001

Susceptibility to chew 132 (550) 067 (444) 00001

Susceptibility to smoke 092 (393) 039 (296) 00001

Reasons to use 278 (707) 251 (592) 00041

N or mat iv e ex pe ct ati ons 3 3 5 ( 10 4 9) 1 6 5 ( 9 62 ) 00001

Data are mean (SD). Estimates are generated from mixed-effects models adjusted for

city, school type, sex, age, and grade*age. High scores on these scales indicate more risk

or less protective.

Table 4: Differences in psychosocial factors associated with tobacco

use, by grade (n=11 642)


(n=6165) (n=5477)

City

C hennai 234% (201267) 89% (67111) 26:1 00001

Delhi 254% (234274) 90% (56124) 28:1 00001

School

Private 108% (50166) 89% (60118) 12:1 05426

Government 287% (267307) 110% (90 130) 26:1 00001

Sex

Girls 201% (176226) 69% (5187) 29:1 00001

Boys 292% (270314) 121% (95147) 24:1 00001

Age (years)

11 166% (144188) 83% (35131) 20:1 00003

12 253% (228278) 63% (4482) 40:1 00001

13 303% (267339) 91% (75107) 33:1 00001

14 321% (259383) 121% (80162) 27:1 00001

Data in parentheses are 95% CI. Estimates generated from mixed-effects models

adjusted for (when not stratified by) city, school type, sex, age, and grade*age.

*Compares prevalence of tobacco use in sixth grade with that in eighth grade.

Table 2: Differences in prevalence of ever use of any kind of tobacco

between sixth-grade and eighth-grade students, by city, type of

school, sex, and age (n=11 642)

7/28/2019 4 Lancet

5/6

Articles

www.thelancet.com Vol 367 February 18, 2006 593

group. Young adolescents are particularly likely tointernalise messages from society.5,8 Despite policies tocontrol tobacco,4 messages in a rapidly changing Indiamay be increasingly pro-tobacco, with greater exposure tomedia from other countries, smoking in Bollywoodmovies, and images via the internet. For example, theCigarettes and Other Tobacco Products Act was passed inIndia in 2003, prohibiting all direct and indirectadvertising of tobacco products, smoking in publicplaces, sales of tobacco products to people younger than18 years, and sales of tobacco products near educationalinstitutions.4 In response, one tobacco company(Godfrey-Philips India) positioned air-conditioned lorries(Mobile Smoking Lounges) outside major attractions,

such as a sports stadium and shopping malls, in fourmajor cities in India, including Delhi, so smokers can sitin the lounge and smoke in comfort. Thus, althoughIndia is a leader in global tobacco control, the tobaccoindustry has also developed methods to counteract orameliorate the effects of policy changes, and theseactions may be of particular interest to young adolescentsas they begin to explore the adult world.4,5

Notably, 108% of the sixth-grade students in ourcohort had ever smoked cigarettes, compared with 42%(in Delhi) and 55% (in Tamil Nadu, the state whereChennai is located) of students in the eighth-to-tenthgrades who participated in the Global Youth TobaccoSurvey in 2001.19 Likewise, 67% of sixth-grade studentsin our study were current users of tobacco, compared

with 4% and 71% of students in the Global YouthTobacco Survey in 2001. Even with the increasesuggested by the recent Indian Cancer Association data,9

our younger sixth-grade cohort of students was usingtobacco at similar or substantially higher rates thanstudents who were 24 years older. Clearly, surveys oftobacco use in students should begin before eighthgrade (about age 13 years) if a true measure of early-onset rates is to be achieved.

The only demographic group in which sixth-gradestudents did not significantly differ from eighth-gradestudents was among those in private schools. However,the private schools had very little age dispersion, withsmall sample sizes in one of the grades for ages 11 yearsand younger, 13 years, and 14 years and older. The onlyage group with substantial numbers of students in bothgrades was 12-year-olds, in whom the rate of tobacco usewas nearly four times greater for sixth-grade than foreighth-grade students. Thus, private schools might alsobe facing increased use in younger children, althoughthe data from this study are inconclusive.

Other explanations for these outcomes include under-reporting or over-reporting by one grade level, lack ofunderstanding of the survey items by the sixth-gradestudents, or differential school drop-out of tobacco usersfrom sixth to eighth grades. Methods were used toensure confidentiality in data collection; these methodshave yielded valid responses in previous work.6,20 An

extensive process of survey development was under-taken to ensure that sixth-grade students understood theexact meanings of the questions on the survey.Moreover, the consistency of the data across differentpopulations, and the reliability and predictive validity ofthe measures, suggest that confusion about thequestions or over-reporting or under-reporting by onegrade level would not account for the degree ofdifferences noted between grades. Additionally, attritionrates are low (yearly, about 1012%) between sixth andeighth grade in the Project MYTRI schools, comparedwith rates before sixth grade, and are not large enough toaccount for the differences seen between grades. Forexample, since 218% of sixth-grade and 6% of eighth-

grade students have ever used tobacco (table 1), theneven if disproportionately more tobacco users droppedout than non-users, about half of the tobacco userswould have to drop out every year from sixth to eighthgrade (and there could be no new users) in order tomatch the lower rates of eighth-grade students. Clearly,since this is a cross-sectional study, longitudinal data areneeded to confirm the sustained increase in tobacco userates in this cohort over time.

Limitations of this study include its cross-sectionaldesign, the need for replication and follow-up of thecohort until adulthood, and the absence of physiologicaldata as another measure of tobacco-use behaviour. Thesample of schools was not randomly selected from thepopulation, but was representative of the mix of types of

schools in these cities and does provide a sample ofstudents of lower to higher socioeconomic status andboth sexes.

The difference in rates of tobacco use between the sixthand eighth grades, and the replication of this differencein two cities, in government schools, and for girls andboys, strongly suggests that sixth-grade students in urbanIndia use tobacco at two to four times the rate that eighthgraders do. Of particular concern are older sixth-gradestudents in government schools, who already reporthaving used tobacco at high rates. These findings mightindicate the initial wave of a large increase in tobacco usein India, which is alarming and warrants confirmationand early intervention in young students.

ContributorsK S Reddy is principal investigator of Project MYTRI in India. Heworked on all aspects of survey development and design, data collectionprotocols, data management, and data analysis. He read, edited, andwrote portions of the manuscript. C L Perry is principal investigator ofProject MYTRI and is responsible for the scientific integrity of thestudy. She worked on all aspects of survey development and design, datacollection protocols, data management, and data analysis. She wrote theresearch article with input from the co-authors. M H Stigler is projectdirector of Project MYTRI in Minnesota and is responsible for theongoing management of the scientific components of the project. Sheworked on all aspects of survey development and design, data collectionprotocols, data management, and data analysis. She did data analysesfor this paper and wrote sections pertaining to the study methods andresults. M Arora is project director of Project MYTRI in India and wasresponsible for training the study staff, coordinating the ongoingmanagement of all staff in India, and implementation of all aspects of

7/28/2019 4 Lancet

6/6

Articles


the project. She worked on all aspects of survey development anddesign, data collection protocols, data management, and data analysis.She read, edited, and wrote portions of the paper.

Conflict of interest statement

We declare that we have no conflict of interest.

Acknowledgments

The research was funded by a grant from the Fogarty InternationalCenter (R01TW05952-01; C L Perry, principal investigator).

References1 WHO. The millennium development goals and tobacco control.

Geneva: World Health Organization, 2005.

2 Ezzati M, Lopez AD. Estimates of global mortality attributable tosmoking in 2000. Lancet2003; 362: 84752.

3 Stewart BW, Kleihues P, eds. World cancer report. Lyon:IARC Press, 2003.

4 Reddy KS, Gupta PC. Tobacco control in India. New Delhi:

Ministry of Health and Family Welfare, Government of India,2004.

5 Perry CL. The tobacco industry and underage teen smoking:tobacco industry documents from the Minnesota litigation.Arch Pediatr Adolesc Med1999; 153: 93541.

6 US Department of Health and Human Services. Preventingtobacco use among young people: a report of the Surgeon General.Atlanta: Centers for Disease Control and Prevention, Office onSmoking and Health, 1994

7 Pierce JP, Gilpin EA. A historical analysis of tobacco marketing andthe uptake of smoking by youth in the US: 18901977.Health Psychol1995; 14: 50008.

8 Pierce JP, Lee L, Gilpin EA. Smoking initiation by adolescent girls:an association with targeted advertising.JAMA 1994; 271:60811.

9 Sharma S. Smoking to look cool, thin. Hindustan Times (Delhi),March 2, 2005.

10 Perry CL, Stigler MH, Arora M, Reddy KS. Prevention intranslation: tobacco use prevention in India.Health Promot Pract(in press).

11 Centers for Disease Control and Prevention. Global Youth TobaccoSurvey (GYTS), Delhi, India fact sheet. http://www.cdc.gov/tobacco/global/gyts/factsheets/2001/India_delhi_factsheet.htm(accessed July 29, 2005).

12 Warren CW, Riley L, Asma S, et al. Tobacco use by youth: asurveillance report from the Global Youth Tobacco Survey project.Bull World Health Organ 2000; 78: 86876.

13 Mishra A, Arora M, Stigler MH, et al. Indian youth speak abouttobacco: results of focus group discussions with school students.Health Educ Behav2005; 32: 36379.

14 Botvin GJ, Botvin EM, Baker E, Dusenbury L, Goldberg CJ. Thefalse consensus effect: predicting adolescents tobacco use fromnormative expectations. Psychol Rep 1992; 70: 17178.

15 Eckhardt L, Woodruff SI, Elder JP. A longitudinal analysis ofadolescent smoking and its correlates.J Sch Health 1994; 64: 6772.

16 Komro KA, Perry CL, Munson KA, Stigler MH, Farbakhsh K.Reliability and validity of self-report measures to evaluate drug andviolence prevention programs.J Child Adolesc Substance Abuse2004; 13: 1751.

17 Pierce JP, Coi WS, Gilpin EA, Farkas AJ, Merrit RK. Validation ofsusceptibility as a predictor of which adolescents take up smokingin the US. Health Psychol1996; 15: 35561.

18 Raudenbush SW, Bryk AS. Hierarchical linear models: applicationsand data analysis methods (2nd edn). Thousand Oaks, CA: SagePublications, 2002.

19 Centers for Disease Control and Prevention (CDC). IndiaTamilNaduStandard 8-0 Global Youth Tobacco Survey (GYTS) factsheet. http://www.cdc.gov/tobacco/global/gyts/factsheets/2001/India_TamilNadu_factsheet.htm (accessed July 29, 2005).

20 Komro KA, Kelder SH, Perry CL, Klepp K-I. Effects of a salivapipeline procedure on adolescent self-reported smoking behavior andyouth smoking prevention outcomes. Prev Med1993; 22: 85765.

4 lancet

Documents