	Table of contents

	Current issue

	Search journal

	Archived issues

	NZMJ Obituaries

	Classifieds

	Hotline (free ads)

	How to subscribe

	How to contribute

	How to advertise

	Contact Us

	Copyright

	Other journals

Journal of the New Zealand Medical Association, 14-December-2007, Vol 120 No 1267

Parental and adolescent smoking: does the association vary with gender and ethnicity?

Robert Scragg, Marewa Glover

Abstract

Aim Determine whether parental smoking is a consistent risk factor for adolescent smoking in a multi-ethnic sample, and whether maternal and paternal effects combine additively or multiplicatively.

Methods Annual national cross-sectional surveys (2002–2004 combined) with multi-ethnic sample of 91,219 Year 10 students from New Zealand who answered a questionnaire on personal and parental smoking.

Results Maternal smoking and paternal smoking were associated separately with increased risk of daily adolescent smoking in all ethnic groups except paternal smoking in Asian youth. The relative risk of adolescent daily smoking (adjusted for age and sex) was significantly higher for maternal only smoking compared with paternal only smoking in each ethnic group: Asians 5.50 (95% CI: 3.55–8.52), Europeans 1.38 (1.26–1.52), Pacific Islanders 1.38 (1.10–1.73), and Māori 1.10 (1.00–1.21). The excess maternal effect varied inversely with smoking prevalence. The net effects of maternal and parental smoking are additive among European, Māori, and Pacific Island students, but multiplicative in Asian. Overall, about 40% of adolescent daily smokers could be attributed to parental smoking.

Conclusions These results show that parental smoking is a consistent risk factor for adolescent smoking in all ethnic groups. Prevention strategies targeted at parents may help limit the uptake of smoking by adolescents.

Parental smoking was identified as a risk factor for adolescent smoking in the 1950s.1–3 Since then, numerous studies of the association between parental and adolescent smoking have been carried out, with inconsistent findings;4 and major questions remain unanswered.

For example, a number of studies have shown that the parental influence is stronger in families where both parents smoke rather than one.2–3,5–12 However, it is unclear whether maternal and paternal effects are additive or whether they interact to increase the risk of adolescent smoking more than the sum of their net effects. Three previous studies have reported additive effects,8,10,13 but this finding is not consistent.14

On balance, maternal smoking has been found to have a stronger effect than paternal smoking,14–17 but it is unclear whether the maternal effect is equal in boys and girls,10 or stronger in girls16–18 such that same-sex parental effects predominate.19–21

Most studies have been of single ethnic groups, and it is unclear whether parental effects are consistent across ethnicities. In the US, adult smoking has been associated with adolescent smoking in whites and Asians but not in blacks and Hispanics;22 while paternal smoking was associated with smoking by white and Latino adolescents, maternal smoking with Asian adolescent smoking, but no association was observed in blacks.23 We have previously shown that parental smoking is a risk factor in all main ethnicities in a 2001 national survey of New Zealand children.24

We have examined the above questions by combining data from the 2002–2004 New Zealand national surveys to provide a large multi-ethnic sample of over 90,000 adolescents. This sample, more than five times larger than any previous study of parental and adolescent smoking,4,25 provided an opportunity to investigate gender-specific smoking effects within each of four main ethnicities (European, Māori, Pacific Islander [mostly of Samoan, Tongan, Niuean, or Cook Islands origin], and Asian [mostly of Chinese, Indian subcontinent, or Korean origin]), to see if parental effects were similar or varied across ethnic groups which have a 10-fold variation in adolescent smoking prevalence, and to assess whether maternal and paternal effects combined in an additive or interactive manner.

Methods

National surveys of tobacco smoking by Year 10 (4th form) students have been carried out yearly since 1999.26 The current paper reports data from the 2002–2004 surveys. In each year, all New Zealand schools with Year 10 students were invited to participate in the survey by administering a short questionnaire to their Year 10 students in November. The annual school response rate was 67.3% in 2002 (n = 309), 66.1% in 2003 (n = 312), and 64.7% in 2004 (n = 319).

The Ethics Committee of the Ministry of Health in Auckland granted a waiver of the formal review and consenting processes. School principals gave permission for teachers to supervise students who completed the anonymous questionnaire in class. To maintain confidentiality, teachers did not examine questionnaires for completeness.

Students answered a two-page questionnaire, which included questions on age, sex, and ethnicity (self-assigned). The questionnaire asked whether the student had ever smoked a cigarette (even just a few puffs). Those who answered “no” were classified as never smokers. Those who answered “yes” were queried about the frequency of their current smoking (at least once a day, at least once a week, at least once a month, less often, never).

Students were asked whether their mother and/or father were current smokers, with the question “Which of these people smoke?” (with instructions to tick any of the following responses that apply: mother, father, older brother or sister, best friend, none of these). For ethnic-specific analyses, because students could choose more than one ethnic group, a priority system was used to classify any student choosing Māori, then any Pacific student, followed by any Asian student, followed by European.

The total number of completed questionnaires returned by schools during the 3-year period was 99,063 (30,972 in 2002, 34,812 in 2003, 33,279 in 2004), out of 140,721 on school rolls (70.4% student response).

Statistical analyses were restricted to 91,219 students who were 14 and 15 years old, with known sex, ethnicity (European, Māori, Pacific, and Asian), student smoking status and parental smoking status; after excluding: 2907 with age 13 years, 16 years or missing; 268 with missing sex; 971 with missing ethnicity; 784 with missing student smoking status; 1386 with missing parent smoking status; and 1528 of “Other” ethnicity.

Statistical analyses were made using SUDAAN (Release 9.0.1, 2005) which corrects standard errors and confidence intervals for any design effect from clustering of students by school. The CROSSTAB procedure was use to calculate prevalences and Mantel-Haenszel adjusted relative risks. Confidence intervals for excess prevalences were calculated from the pooled variance.27 The population attributable risk was calculated by estimating the attributable proportion for the exposed cases within each exposure category.28,29

Results

Prevalence of adolescent and parental smoking—The sample analysed comprised 46,400 girls (30,569 Europeans, 8632 Māori, 3103 Pacific Islanders, 4096 Asians) and 44,819 boys (29,414 Europeans, 7839 Māori, 3262 Pacific Islanders, 4304 Asians). The prevalence of daily smoking was higher in girls (13.6%; 95% confidence interval [CI] 12.5%–14.7%) compared to boys (9.3%; 95% CI 8.5%–10.0%). There was a more than 8-fold variation in daily smoking prevalence among girls, from Māori 32.4%, Pacific 16.3%, European 9.4%, to Asian 3.7%; and 3-fold variation among boys from Māori 17.5%, Pacific 11.8%, European 7.3%, to Asian 5.7% (Figure 1).

Figure 1. Prevalence (95% CI) of daily smoking by gender and ethnicity

The prevalence of parental smoking varied with student ethnicity and sex of the parent (Table 1). Parents of Māori students were most likely to smoke (64.4%), followed by Pacific Island (47.5%), and European (34.6%) parents, with Asian parents least likely to smoke (30.6%). There was a 6-fold variation in the sex-ratio of smoking parents between ethnic groups, with the ratio of maternal-to-paternal smokers ranging from a high of 1.48 for Māori, down to 0.97 for European, and 0.95 for Pacific, and right down to 0.24 for Asian parents.

Table 1. Distribution of parental smoking status among male and female students, by ethnicity

Parental smoking	European	Māori	Pacific*	Asian†	Total
Both smoke Mother only Father only Neither (n)	11.4% 11.1% 12.2% 65.4% (59,983)	30.2% 20.6% 13.6% 35.6% (16,471)	16.8% 12.5% 18.2% 52.5% (6365)	4.5% 1.9% 24.2% 69.4% (8400)	14.5% 12.0% 14.0% 59.5% (91,219)

*Mostly of Samoan, Tongan, Niuean, or Cook Island origin; †Mostly of Chinese, Indian subcontinent, or Korean origin.

Risk of adolescent smoking associated with parental smoking—The prevalence of daily youth smoking, along with the excess daily smoking prevalence (compared with students of non-smoking parents), by parental smoking status, is shown in Table 2, by sex and ethnicity. Maternal smoking and paternal smoking, alone, were each associated with a significant (p<0.05) excess smoking prevalence in all student sex-ethnic groups aside from paternal smoking among Pacific and Asian students of either sex.

Relative risks of student daily smoking associated with parental smoking, by sex and ethnicity of students, adjusted for age, are shown in Table 3. As reported in Table 2, maternal only and paternal only smoking were each associated with a significantly increased relative risk in all student subgroups except for paternal only smoking in Pacific and Asian students of either sex.

When male and female students in the latter two ethnic groups were combined, the relative risk associated with paternal only smoking (adjusted for age and sex) was significantly increased compared with neither parent smoking among Pacific students (1.29; 95% CI: 1.05–1.58) but not in Asian students (1.22; 95% CI: 0.91–1.64). Thus, paternal only smoking was associated with significantly increased risk of daily student smoking in all ethnic groups except Asian.

Risk of adolescent smoking by parent gender—Maternal-only smoking had a stronger effect than paternal-only smoking. When students in these two parental smoking groups were compared, the relative risk of student daily smoking (adjusted for age and sex) was significantly higher for maternal only smoking compared with that for paternal only smoking in each ethnic group: Asians 5.50 (95% CI: 3.55–8.52), Europeans 1.38 (95% CI: 1.26–1.52), Pacific Islanders 1.38 (95% CI: 1.10–1.73), and Māori 1.10 (95% CI: 1.00, 1.21; p=0.050).

Figure 2 shows that the increased effect from maternal smoking (compared with paternal smoking) varied inversely with the ratio of maternal to paternal smokers in each ethnic group, so that the increased maternal effect is strongest in Asian students where maternal smoking is rare and weakest in Māori students where maternal smoking is most common.

The data in Tables 2 and 3 provide limited support for gender-specific effects on students associated with maternal smoking. The absolute effect of mother only smoking was larger in girls than boys among European (11.8% vs 7.6%) and Māori (15.9% vs 8.0%) students, with confidence intervals of comparison groups not overlapping, but not different between female and male Pacific (10% vs 5.4%) and Asian (16.4% vs 15.9%) students; while, the absolute effect of father only smoking was the same for girls and boys in each ethnic group (Table 2).

The same pattern is seen in Table 3, where relative risks of student daily smoking associated with father only smoking are very similar for girls and boys within each ethnic group; while only among European students was the relative risk associated with mother only smoking higher for girls compared to boys (Breslow-Day test for homogeneity, p-value <0.05).

Table 2. Daily smoking prevalence, and excess daily smoking prevalence—by sex, ethnicity, and parental smoking status

Ethnicity	Parental smoking	Girls			Boys
Ethnicity	Parental smoking	N	Daily smoking prevalence % (95% CI)	Excess prevalence %	N	Daily smoking prevalence % (95% CI)	Excess prevalence %
European	Both smoke Mother only Father only Neither	3536 3376 3819 19,838	25.6 (23.8–27.4) 16.6 (15.2–18.1) 11.7 (10.4–12.9) 4.8 (4.3–5.3)	20.8 (19.0–22.7) 11.8 (10.3–13.4) 6.9 (5.5–8.2) 0	3297 3257 3472 19,388	19.8 (18.3–21.2) 11.8 (10.5–13.2) 8.9 (7.8–10.1) 4.2 (3.8–4.6)	15.6 (14.1–17.1) 7.6 (6.3–9.0) 4.7 (3.6–6.0) 0
Māori	Both smoke Mother only Father only Neither	2699 1802 1176 2955	44.7 (42.3–47.0) 35.6 (33.1–38.0) 31.5 (28.3–34.7) 19.7 (17.9–21.6)	25.0 (21.9–27.9) 15.9 (12.8–18.9) 11.8 (8.0–15.4) 0	2270 1587 1067 2915	26.0 (23.5–28.5) 18.4 (16.2–20.7) 17.5 (15.0–20.0) 10.4 (9.0–11.8)	15.6 (12.8–18.4) 8.0 (5.4–10.6) 7.1 (4.3–10.0) 0
Pacific*	Both smoke Mother only Father only Neither	508 383 599 1613	27.8 (23.9–31.6) 21.9 (17.4–26.5) 15.0 (11.3–18.8) 11.9 (10.0–13.8)	15.9 (11.5–20.2) 10.0 (5.1–15.0) 3.1 (-1.1–7.4) 0	562 410 559 1731	24.4 (20.6–28.1) 13.2 (9.9–16.4) 10.4 (8.0–12.8) 7.8 (6.1–9.5)	16.6 (12.5–20.7) 5.4 (1.7–9.1) 2.6 (-0.4–5.6) 0
Asian†	Both smoke Mother only Father only Neither	157 70 1039 2830	35.7 (27.8–43.6) 18.6 (10.2–27.0) 2.1 (1.2–3.0) 2.2 (1.5–2.8)	33.5 (25.6–41.5) 16.4 (8.0–24.8) -0.1 (-1,2–1.1) 0	219 89 993 3003	41.1 (33.0–49.2) 19.1 (9.3–28.9) 4.5 (2.9–6.2) 3.2 (2.1–4.3)	37.9 (29.8–46.1) 15.9 (6.1–25.8) 1.3 (-0.6–3.4) 0

*Mostly of Samoan, Tongan, Niuean, or Cook Islands origin; †Mostly of Chinese, Indian subcontinent, or Korean origin.

Table 3. Relative risk of daily smoking associated with parental smoking status, by sex and ethnicity

Ethnicity	Parental smoking	Girls	Boys
Ethnicity	Parental smoking	Relative Risk (95% CI)*	Relative Risk (95% CI)*
European	Both smoke Mother only Father only Neither	5.35 (4.83–5.91) 3.46 (3.08–3.87) 2.44 (2.17–2.74) 1.00	4.70 (4.22–5.24) 2.83 (2.49–3.21) 2.14 (1.89–2.42) 1.00
Māori	Both smoke Mother only Father only Neither	2.26 (2.07–2.45) 1.80 (1.63–1.99) 1.60 (1.43–1.80) 1.00	2.50 (2.20–2.85) 1.77 (1.51–2.08) 1.69 (1.42–2.01) 1.00
Pacific**	Both smoke Mother only Father only Neither	2.33 (1.95–2.77) 1.84 (1.41–2.41) 1.26 (0.97–1.63) 1.00	3.13 (2.41–4.07) 1.68 (1.19–2.36) 1.33 (0.97–1.81) 1.00
Asian†	Both smoke Mother only Father only Neither	16.60 (11.52–23.92) 8.62 (4.96–14.97) 0.99 (0.61–1.59) 1.00	12.61 (9.00–17.66) 5.90 (3.30–10.56) 1.38 (0.95–2.01) 1.00

*Adjusted for age; **Mostly of Samoan, Tongan, Niuean, or Cook Islands origin; †Mostly of Chinese, Indian subcontinent, or Korean origin.

Are parental effects additive or multiplicative?—For students with both parents smoking, the excess prevalence was similar to the sum of the individual excess prevalence for maternal only smoking and paternal only smoking among European and Māori students of both sexes and Pacific Island girls, but larger among male and female Asian students and male Pacific Island students (Table 2).

This pattern is repeated with the relative risks in Table 3, where the increase in the relative risk for students exposed to both parents is more than the sum of the increases for mother only smoking and father only smoking for Asian students and male Pacific Island students. However, in logistic regression models which contained maternal smoking, paternal smoking and their interaction term, the beta-coefficients for the interaction term were only positive and significant in Asian female and male students (p=0.0063). These results indicate that the effects of maternal and parental smoking are additive among European, Māori, and Pacific Island students, but multiplicative in Asian.

Figure 2. Relation between ratio of maternal to paternal smokers (x) and ratio of effect from maternal and paternal smoking (y)

The proportions of student daily smokers attributable to parental smoking are shown in Table 4. Overall, 43% of female student daily smokers, and 41% of male, could be attributed to parental smoking. The proportion attributable to maternal smoking is higher than for paternal smoking.

For girls, 26% of daily smokers are attributable to maternal smoking compared to 15% for paternal smoking; and among boys the respective proportions are 22% and 15%. This pattern is consistent within each sex-ethnic subgroup, and is primarily due to the greater effect of maternal smoking, than paternal smoking, in increasing student smoking as described above.

Discussion

We have found in a large multi-ethnic cross-sectional sample of more than 90,000 adolescents that parental smoking is a risk factor for adolescents of all ethnic groups; and that maternal smoking has a greater effect than paternal smoking on the risk of adolescent smoking.

A major strength of our analysis is the large sample size, more than 5 times larger than any previous study, which has ensured sufficient statistical power to detect parental smoking effects in all ethnic subgroups. Previous studies which have examined the effect of parental smoking across ethnicities have had small sample sizes. For example, small sample size is a likely explanation for the lack of a consistent parental smoking effect in 211 Hispanic and 80 African-American students studied by Sussman and colleagues;22 while Landrine and colleagues did not observe a parental smoking effect in a sample of only 514 African-Americans.23

Both maternal smoking and paternal smoking were associated separately with increased risk of adolescent smoking in all ethnic groups except for paternal smoking among Asian students (Tables 2 and 3).

Our finding of a stronger maternal smoking effect, than paternal, on the risk of adolescent smoking is consistent with most previous studies14–17 but not all.10 It is possible that previous studies have not detected a weaker paternal effect because of their smaller sample sizes.15 A new finding from our data is that the excess maternal smoking effect is inversely associated with the ratio of maternal to paternal smokers (Figure 2).

In populations where maternal smoking is uncommon (e.g. Asians), those mothers who smoke appear to have a much greater effect on the risk of their children smoking than in populations where maternal smoking is common (e.g. Māori). So great is the maternal smoking effect among Asian adolescents, that there was a multiplicative interaction between maternal and paternal smoking effects. However, among all other ethnic groups (European, Māori, Pacific Islander), maternal and paternal smoking effects were additive, consistent with previous reports.8,10,13

We found only limited support for gender-specific effects (Table 2). Maternal smoking had a stronger effect in girls than boys among European and Māori students, but did not differ between girls and boys among Pacific and Asian students; while the absolute effect of paternal smoking was the same for girls and boys in all ethnic groups. Our findings contrast with earlier studies carried out in the 1960s to 1980s which reported gender-specific effects.19–21 The reasons for this are unclear, but changing parental role models with increased sharing of household duties by both parents is a possible explanation.

Further, our study highlights the importance of reporting both absolute and relative effects. The excess (or absolute) effect of parental smoking on adolescent smoking was similar for European, Māori, and Pacific students (Table 2). In contrast, the relative effect from parental smoking varied between these three ethnic groups, being highest in European students and lowest in Māori students (Table 3), because of varying adolescent smoking prevalences among the reference group (students with neither parent smoking), which were low in European and highest in Māori (Table 2).

Our study has strengths and weaknesses. Its major strengths are the very large sample size, thus ensuring the study has high statistical power, and varied ethnic composition. However, the cross-sectional study design does limit the inferences that can be drawn around the causal sequence, although it is reasonable to assume that parental smoking precedes student smoking in all or most instances.

The overall response rate is only moderate, after allowing for school and student non-response, and hence the results cannot be extrapolated with certainty to all Year 10 students in New Zealand. However, the 70% student response rate suggests that the results are likely to be representative of participating schools.

The collection of parental smoking data from students may have resulted in errors in measuring parental smoking status, but if this was non-differential, it is likely to have attenuated the effect of parental smoking. In particular, some parents who were classified as non-smokers are likely to have been ex-smokers, and their presence in the non-smoking parental group will have biased effect estimates towards the null. Many comparisons are in this analysis, and it is possible that some significant results occurred by chance, although we consider this is unlikely because of the large sample sizes in this study.

Our attributable risk percentages for parental smoking (43% in girls and 41% in boys, Table 4) indicate that parental smoking is a major cause of adolescent smoking, which, in combination with other factors under parental control (e.g. allowing smoking in the home), explains a similar proportion of adolescent daily smoking as does peer smoking.30

Our results are similar to those reported by Kandel and Wu who found that maternal smoking explained 60% of smoking by girls and 30% by boys.14 These findings suggest that prevention strategies targeted at parents may help limit the uptake of smoking by adolescents.

Competing interests: None.

Author information: Robert Scragg, Associate Professor in Epidemiology; Marewa Glover, Research Fellow; School of Population Health, University of Auckland, Auckland

Acknowledgements: The survey was carried out by Action on Smoking and Health (ASH). The New Zealand Ministry of Health provided funds. The manuscript was initiated and analysed by the authors.

Correspondence: Assoc Prof Robert Scragg, Epidemiology & Biostatistics, School of Population Health, University of Auckland, Private Bag 92019, Auckland Mail Centre, Auckland 1142, New Zealand. Fax: +64 (0)9 3737 503; email: r.scragg@auckland.ac.nz

References:

Bothwell P. The epidemiology of cigarette smoking in rural children. Medical Officer. 1959;102:125–32.
Horn D, Courts F, Taylor R, Solomon E. Cigarette smoking among high school students. Am J Publ Health. 1959;49:1497–511.
Salber E, MacMahon B. Cigarette smoking among high school students related to social class and parental smoking habits. Am J Publ Health. 1961;51:1780–9.
Avenevoli S, Merikangas KR. Familial influences on adolescent smoking. Addiction 2003;98(suppl 1):1–20.
Bewley BR, Bland JM. Academic performance and social factors related to cigarette smoking by schoolchildren. Br J Prev Soc Med. 1977;31:18–24.
Lemin B. Smoking in 14-year-old school children. Int J Nurs Studies. 1967;4:301–7.
Morison M, Kiryluk S, Swan AV. Health education and cigarette smoking. Can Med Assoc J. 1964;91:49–56.
Nolte AE, Smith BJ, O’Rourke T. The relative importance of parental attitudes and behaviour upon youth smoking behaviour. J School Health. 1983;53:264–71.
Palmer AB. Some variables contributing to the onset of cigarette smoking among junior high school students. Soc Sci Med. 1970;4:359–66.
Peterson AV, Leroux, BG, Bricker J, et al. Nine-year prediction of adolescent smoking by number of smoking parents. Addict Behav. 2006;31:788–801.
Szabo E, White V, Hayman J. Can home smoking restrictions influence adolescents’ smoking behaviors if their parents and friends smoke? Addict Behav. 2006;31:2298–303.
Williams AF. Personality and other characteristics associated with cigarette smoking among teenagers. J Health Soc Behav. 1973;14:374–80.
Rossow I, Rise J. Concordance of parental and adolescent health behaviours. Soc Sci Med. 1994;38:1299–305.
Kandel DB, Wu P. The contributions of mothers and fathers to the intergenerational transmission of cigarette smoking in adolescence. J Res Adolescence. 1995;5:225–52.
Melby JN, Conger RD, Conger KJ, Lorenz FO. Effects of parental behaviour on tobacco use by young male adolescents. J Marriage Family. 1993;55:439–54.
Øygard L, Klepp K-I, Tell GS, Vellar OD. Parental and peer influences on smoking among young adults: ten-year follow-up of the Oslo youth study participants. Addiction. 1995;90:561–9.
Swan AV, Creeser R, Murray M. When and why children first start to smoke. Int J Epidemiol. 1990;19:323–330.
Byrne DG, Byrne AE, Reinhart MI. Psychosocial correlates of adolescent cigarette smoking: personality or environment. Aust J Psychol. 1993;45:87–95.
Beaglehole R, Eyles E, Harding W. Cigarette smoking habits, attitudes, and associated factors in adolescents. N Z Med J. 1978;87:239–42.
Burchfiel CM, Higgins MW, Keller JB, et al. Initiation of cigarette smoking in children and adolescents in Tecumseh, Michigan. Am J Epidemiol. 1989;130:410–15.
Levitt E, Edwards J. A multivariate study of correlative factors in youthful cigarette smoking. Dev Psychol. 1970;2:5–11.
Sussman S, Dent CW, Flay BR, et al. Psychosocial predictors of cigarette smoking onset by white, black, Hispanic and Asian adolescents. MMWR. 1987;36 (suppl 4):11S–16S.
Landrine H, Richardson JL, Klonoff EA, Flay B. Cultural diversity in the predictors of adolescent cigarette smoking: the relative influence of peers. J Behavioural Med. 1994;17:331–46.
Scragg R, Laugesen M, Robinson E. Parental smoking and related behaviours influence adolescent smoking: results from the 2001 New Zealand national survey of 4th form students. N Z Med J. 2003;116(1187). http://www.nzma.org.nz/journal/116-1187/707
Conrad KM, Flay BR, Hill D. Why children start smoking cigarettes: predictors of onset. Br J Addict. 1992;87:1711–24.
Scragg R. Report of 1999-2005 National Year 10 Smoking Surveys: prepared for ASH. Auckland: ASH, 2006. http://ash.org.nz/system/_doc.php?id=860
Colton T. Statistics in medicine. Boston: Little, Brown and Co, 1974.
Rockhill B, Newman B, Weinberg C. Use and misuse of population attributable fractions. Am J Publ Health. 1998;88:15–19.
Rothman K. Modern Epidemiology. Oxford: OUP, 1986.
Scragg R, Laugesen M. Influence of smoking by family and best friend on adolescent tobacco smoking: results from the 2002 New Zealand national survey of Year 10 students. ANZ J Publ Health. 2007;31:217–23.