bs_bs_banner
doi:10.1111/jpc.12667
ORIGINAL ARTICLE
Independent and joint effects of prenatal maternal smoking and maternal exposure to second-hand smoke on the development of adolescent obesity: A longitudinal study Liang Wang,1 Hadii M Mamudu,2 Arsham Alamian,1 James L Anderson1 and Billy Brooks1 1 Department of Biostatistics and Epidemiology and 2Department of Health Services Management and Policy, College of Public Health, East Tennessee State University, Johnson City, Tennessee, United States
Aim: To examine associations of prenatal maternal smoking and second-hand smoke (SHS) exposure with the development of adolescent obesity. Methods: Longitudinal data (1991–2007) from National Institute of Child Health and Human Development Study of Early Child Care and Youth Development involving mothers that smoked and or exposed to SHS during the year before birth were analysed. Adolescent obesity in ages 12.0–15.9 years was defined as a BMI ≥ 95th percentile. Generalised estimating equations (GEE) were used for the analyses. Results: Obesity was more prevalent among adolescents whose mothers smoked or had SHS exposure than those that did not smoke or exposed to SHS. After adjusting for maternal and child factors, GEE models showed that odds of adolescent obesity increased with prenatal maternal smoking (OR = 1.57, 95% CI = 1.03–2.39) and SHS exposure (OR = 1.53, 95% CI = 1.04–2.27). The odds for obesity increased more than two times among adolescents exposed to both maternal smoking and SHS (OR = 2.10, 95% CI = 1.24, 3.56) compared with those without exposure. Additionally, not breastfeeding, maternal obesity, and longer screen viewing hours per day were associated with increased odds of obesity. Conclusions: There is possibly a long-term joint effect of prenatal maternal smoke (smoking and SHS) exposure on obesity among adolescent offspring, and the effect is independent of birthweight. These findings suggest that adolescent obesity could possibly be curtailed with the development and promotion of smoking cessation programmes for families during the year before birth. Key words:
adolescent obesity; longitudinal study; maternal smoking.
What is already known on this topic
What this paper adds
1 Prenatal environments may have long-term effects on health outcomes of children. 2 Prenatal maternal smoking is associated with increased risk of obesity in childhood, but not consistently evident in adolescence. 3 Few longitudinal studies have examined the long-term effect of direct cigarette smoking and second-hand smoke during the prenatal period on the risk of adolescent obesity.
1 Prenatal exposure to cigarette smoke has negative health outcomes on offspring during adolescence. 2 Adolescents whose mothers were exposed to direct cigarette smoking and second-hand smoke during the year before birth were at an increased risk of obesity. 3 Incorporating prenatal smoking cessation in tobacco control programmes could potentially improve the health status.
In 2009–2010, the US Centers for Disease Control and Prevention (CDC) estimated that about 36% of adults and 17% of adolescents were obese.1,2 The rate of adult obesity is expected to reach 42% by 2030.3 Obesity is a risk factor for several diseases, including cardiovascular disease, diabetes and cancer.4,5 Additionally, obesity imposes $147 billion of economic costs in the United States alone, which is about one-tenth of Correspondence: Dr Liang Wang, Department of Biostatistics and Epidemiology, East Tennessee State University, PO Box 70259, Johnson City, TN 37614, USA. Fax: 423 439 6491; email: [email protected] Conflict of interest: The authors have none to declare. Accepted for publication 11 May 2014.
908
annual medical expenditures.6 Adolescent obesity is of public health concern because obese adolescents are more likely to become obese adults,7,8 have impaired health-related quality of life,9 problems with school functioning regarding health issues,10 bullied during school-going ages11 and die prematurely in adulthood.7 Therefore, it is important to identify causes of adolescent obesity, especially as the programmes that have focused on adolescent obesity prevention over the past decade have demonstrated little change in obesity prevalence.3,12,13 Although several factors have been identified as risk factors for adolescent obesity,14 it is necessary to study how environmental factors such as smoking and second-hand smoke (SHS) exposure impact the developmental origins of health and diseases.15 Smoking and SHS exposure have negative health
Journal of Paediatrics and Child Health 50 (2014) 908–915 © 2014 The Authors Journal of Paediatrics and Child Health © 2014 Paediatrics and Child Health Division (Royal Australasian College of Physicians).
L Wang et al.
effects.16 The nicotine in tobacco is addictive and affects the appetite of tobacco users17 with strong nutritional implications. Studies have found that maternal behavior that impacts the intrauterine environment is essential to the development of the offspring18,19 and that maternal smoking and SHS potentially impact the weight status of children.20,21 Thus, there is a need to study maternal smoking behavior and/or exposure to SHS because of the implications for the child development, including risk of childhood and adolescent obesity.22,23 Emerging evidence suggests that prenatal environments may have long-term effects on health outcomes of children,24 which imply a possible association between prenatal factors and obesity among offspring. Previous studies have explored association between prenatal maternal smoking and weight of offspring using body mass index (BMI) measurements of groups ranging from newborn to 33 years of age.25–31 The magnitude of the risk and the dose-response for these relationships vary by age group and sex.31 Children exposed to cigarette smoking in utero are more likely to have a lower birthweight than those without exposure22 and are conversely at an increased risk of being obese.23,30 It has been found that SHS is independently associated with reduced anthropometric measurements of the newborn and increased rate of low birthweight infants.32 A cross-sectional study in the United States examining the impact of maternal and paternal smoking during pregnancy and the risk of overweight and obesity observed the strongest association with adult overweight and obesity when both parents smoked, although BMI was based on self-reported height and weight.21 A prospective study in the Netherlands examined only the independent effect of maternal smoking during pregnancy and ‘smoking in the parental house’ on childhood overweight without assessing the joint effect of smoking and SHS exposure on the childhood overweight.20 Thus, little is known about the impact of prenatal maternal smoking or SHS exposure and the joint effects during the year before birth on adolescent obesity using longitudinal data and accurate anthropometric measurements. The Healthy People 2020 in the United States aims to reduce the proportion of adolescent obesity by 10% from 17.9% in 2005–2008,33 which generates the need to investigate sources of obesity among adolescents. It is the purpose of this study to evaluate the association between prenatal maternal smoking status, measured as either direct cigarette smoking or SHS exposure, during the year before birth and the risk of obesity in the offspring during adolescence. This study could inform smoking cessation among potential pregnant women and programmes to improve maternal and child health.
Method Study sample and data set This study utilised data from the National Institute of Child Health and Human Development (NICHD) Study of Early Child Care and Youth Development (SECCYD), a longitudinal study that followed a cohort of children at four continuous phases (1991–1994, 1995–1999, 2000–2004, 2005–2007), from birth through age 15.9 years. Participants were recruited from 24 hospitals in the vicinity of 10 data collection sites across the United States. A more complete description of data collection
Maternal smoking and adolescent obesity
procedures and protocols of the SECCYD can be found elsewhere.34 Families were enrolled in the study upon successful completion of data collection through the 1-month interview. Recruitment was accomplished during the first 11 months of 1991 and 1364 families were enrolled. We restricted the analyses to adolescents whose mothers completed the question regarding maternal smoking and SHS exposure during the year prior to birth and those who had weight and height data between ages 12.0 and 15.9 years. A total of 932 participants were included in the analytic sample. This study was approved by authors’ institutional review board.
Study variables Obese status of adolescents The outcome variable was adolescent obesity between ages 12.0 and 15.9 years, which is defined as a child whose BMI ≥ 95th percentile according to the CDC’s sex- and-age specific growth chart for the year 2000.35 The BMI of adolescents was calculated using measured height and weight obtained using standard protocols and procedures.30,34 Standardised procedures were used to measure height and weight during the interviews by the NICHD SECCYD staff. Height was measured with children standing without shoes, feet together and their backs against a calibrated 7-foot measuring stick. Weight was measured using a physician’s two-beam scale. Scales were calibrated monthly using certified calibration weights. Weight was measured with children in minimal clothing and recorded twice, each time to the nearest 0.25 pound (0.1 kg).
Prenatal maternal smoking The primary independent variables were mother’s smoking behavior or SHS exposure during the year immediately prior to the child’s birth (1990). These variables were assessed retrospectively at the time the child was 24 months old with questions relating to the ‘The Year before My Child Was Born’.36 Prenatal maternal smoking status was determined by asking mothers ‘Which sentence best describes your smoking pattern during the year before your child was born?’ The responses were (i) I did not smoke; (ii) I smoked but stopped before pregnancy began; (iii) I smoked, and stopped during the first 3 months of pregnancy; (iv) I smoked through the first 3 months but stopped before my child was born; and (v) I smoked throughout the year. Maternal SHS exposure was determined with the question, ‘During the year before your child was born, how frequently did you come in contact with family, friends or co-workers who smoked?’ and the responses were (i) never, (ii) sometimes, (iii) often and (iv) always. Because of the small sample size in some subgroups, mother’s smoking status during the 1 year before birth was combined into two categories: did not smoke (i) and smoked (ii–v). Because only two mothers smoked without exposure to SHS using the never versus sometimes/often/always classification, we re-categorised SHS exposure as ‘never/sometimes’ and ‘often/always’ to increase the sample size for this variable. Participants were categorised into four subgroups according to smoking and SHS exposure status: (i) neither (reference group, ‘the lowest risk group’); (ii)
Journal of Paediatrics and Child Health 50 (2014) 908–915 © 2014 The Authors Journal of Paediatrics and Child Health © 2014 Paediatrics and Child Health Division (Royal Australasian College of Physicians)
909
Maternal smoking and adolescent obesity
L Wang et al.
smoking only; (iii) SHS exposure only; and (iv) both smoking and SHS exposure (‘the highest risk group’).
stepwise fashion (Table 4). All data analyses were performed using SAS version 9.2 (SAS Institute Inc., Cary, NC, USA).
Covariates
Results
Maternal information collected at the time when the child was 24 months old included education (less than high school, high school and some college or above), poverty (above poverty line, at or below poverty line), employment (employed, not employed) and family structure (living together, not living together ). Additionally, self-reported maternal body weight in pounds and height in inches were assessed when the child was 15 years of age, and maternal BMI was calculated (weight (pounds)/height (inches2)* conversion factor of 703). Maternal obesity was defined as a BMI ≥ 30 kg/m2. Child information collected at 1 month after birth included breastfeeding (breastfeeding, not breast feeding), birthweight (kilograms) obtained from the medical chart, sex (male, female), race (White, non-White) and gestational age in weeks (calculated with the formula as integer (40*(birthdate − due date)/7). Screen viewing hours (i.e. hours per day watching TV, video, or DVD) were dichotomised into less than 2 h and 2 h or longer.
Table 1 compares participant characteristics in the analytic sample (n = 932) with the excluded sample (n = 432), showing that mothers in the excluded sample were more likely to have less than a high school degree (P < 0.0001) and to not be employed (P = 0.04). Except for sex (49.3% vs. 56.9% for male, P = 0.01), there was no significant differences between the analytic and excluded samples with respect to child factors. Table 2 shows the comparison of maternal and child factors according to the four subgroups of prenatal maternal smoking and SHS exposure. The proportions of mothers who had less than a high school degree (P < 0.0001), were at or below poverty level (P < 0.0001), were not employed (P = 0.006), were not living together with their partners (P < 0.0001), were not breastfeeding (P < 0.0001), and proportions of children who were non-Whites (P = 0.016) and had 2 h or longer screen
Statistical analysis Chi-square tests or Student’s t-tests were used to compare characteristics of participants in the analytic sample (n = 932) with those in the excluded sample (n = 432). Analysis of variance or chi-square tests were used to compare maternal and child factors according to the four subgroups of prenatal maternal smoking and SHS exposure. In this longitudinal study, adolescents’ weight and height were measured at multiple follow-up time points (grades 5, 6, 7, 8 and 15 years of age) from 12.0 to 15.9 years of age. Thus, adolescent obesity assessed at each time point should not be independent, but be correlated. The generalised estimating equation (GEE) was used to address this issue by using ‘working correlation matrix’ to account for the withinsubject correlation of response on dependent variables of different distributions.37 The missing child’s obesity data were assumed to be missing at random, a regular method in longitudinal studies.38 First-order auto-regression was used as the working correlation matrix for performing the GEE to account for correlations among repeated measurements in the pediatric obesity. GEE regression modeling approaches were employed to evaluate the independent and combined effects of prenatal smoking and SHS exposure on the development of adolescent obesity. In the first approach (entitled Model 1 in Table 3), we examined the univariate associations between prenatal maternal smoking and exposure to SHS and the development of adolescent obesity. In the second approach (Model 2), we examined the individual effect of prenatal smoking and SHS exposure in two separate multivariate regression models with adjustment for maternal factors (education, poverty status, family structure, breastfeeding status, employment, gestational age and maternal obesity) and child factors (sex, race and screen viewing hours per day). We added birthweight to Model 3 (the third approach) with adolescent obesity as the outcome to explore its role as a potential mediator. The joint effect of prenatal maternal smoking and maternal SHS exposure were evaluated in a similar 910
Table 1 Comparison of participant factors in the analytic sample versus the excluded sample (n = 1364) Factors
Maternal factors Education§ Less than high school High school Some college or above Poverty§ Below poverty line At or above poverty line Employment§ Not employed Employed Family structure§ Living together Not living together Child factors Birthweight (kg)¶ Sex§ Female Male Race§ White Non-White
Analytic sample† n = 932 (%)
Excluded sample‡ n = 432 (%)
73 (7.8) 189 (20.3) 670 (71.9)
66 (15.3) 98 (22.7) 267 (62.0)
108 (11.7) 813 (88.3)
35 (13.2) 231 (86.8)
281 (30.2) 650 (69.8)
101 (36.7) 174 (63.3)
797 (86.1) 129 (13.9)
235 (86.7) 36 (13.3)
P-value
doi:10.1111/jpc.12667
ORIGINAL ARTICLE
Independent and joint effects of prenatal maternal smoking and maternal exposure to second-hand smoke on the development of adolescent obesity: A longitudinal study Liang Wang,1 Hadii M Mamudu,2 Arsham Alamian,1 James L Anderson1 and Billy Brooks1 1 Department of Biostatistics and Epidemiology and 2Department of Health Services Management and Policy, College of Public Health, East Tennessee State University, Johnson City, Tennessee, United States
Aim: To examine associations of prenatal maternal smoking and second-hand smoke (SHS) exposure with the development of adolescent obesity. Methods: Longitudinal data (1991–2007) from National Institute of Child Health and Human Development Study of Early Child Care and Youth Development involving mothers that smoked and or exposed to SHS during the year before birth were analysed. Adolescent obesity in ages 12.0–15.9 years was defined as a BMI ≥ 95th percentile. Generalised estimating equations (GEE) were used for the analyses. Results: Obesity was more prevalent among adolescents whose mothers smoked or had SHS exposure than those that did not smoke or exposed to SHS. After adjusting for maternal and child factors, GEE models showed that odds of adolescent obesity increased with prenatal maternal smoking (OR = 1.57, 95% CI = 1.03–2.39) and SHS exposure (OR = 1.53, 95% CI = 1.04–2.27). The odds for obesity increased more than two times among adolescents exposed to both maternal smoking and SHS (OR = 2.10, 95% CI = 1.24, 3.56) compared with those without exposure. Additionally, not breastfeeding, maternal obesity, and longer screen viewing hours per day were associated with increased odds of obesity. Conclusions: There is possibly a long-term joint effect of prenatal maternal smoke (smoking and SHS) exposure on obesity among adolescent offspring, and the effect is independent of birthweight. These findings suggest that adolescent obesity could possibly be curtailed with the development and promotion of smoking cessation programmes for families during the year before birth. Key words:
adolescent obesity; longitudinal study; maternal smoking.
What is already known on this topic
What this paper adds
1 Prenatal environments may have long-term effects on health outcomes of children. 2 Prenatal maternal smoking is associated with increased risk of obesity in childhood, but not consistently evident in adolescence. 3 Few longitudinal studies have examined the long-term effect of direct cigarette smoking and second-hand smoke during the prenatal period on the risk of adolescent obesity.
1 Prenatal exposure to cigarette smoke has negative health outcomes on offspring during adolescence. 2 Adolescents whose mothers were exposed to direct cigarette smoking and second-hand smoke during the year before birth were at an increased risk of obesity. 3 Incorporating prenatal smoking cessation in tobacco control programmes could potentially improve the health status.
In 2009–2010, the US Centers for Disease Control and Prevention (CDC) estimated that about 36% of adults and 17% of adolescents were obese.1,2 The rate of adult obesity is expected to reach 42% by 2030.3 Obesity is a risk factor for several diseases, including cardiovascular disease, diabetes and cancer.4,5 Additionally, obesity imposes $147 billion of economic costs in the United States alone, which is about one-tenth of Correspondence: Dr Liang Wang, Department of Biostatistics and Epidemiology, East Tennessee State University, PO Box 70259, Johnson City, TN 37614, USA. Fax: 423 439 6491; email: [email protected] Conflict of interest: The authors have none to declare. Accepted for publication 11 May 2014.
908
annual medical expenditures.6 Adolescent obesity is of public health concern because obese adolescents are more likely to become obese adults,7,8 have impaired health-related quality of life,9 problems with school functioning regarding health issues,10 bullied during school-going ages11 and die prematurely in adulthood.7 Therefore, it is important to identify causes of adolescent obesity, especially as the programmes that have focused on adolescent obesity prevention over the past decade have demonstrated little change in obesity prevalence.3,12,13 Although several factors have been identified as risk factors for adolescent obesity,14 it is necessary to study how environmental factors such as smoking and second-hand smoke (SHS) exposure impact the developmental origins of health and diseases.15 Smoking and SHS exposure have negative health
Journal of Paediatrics and Child Health 50 (2014) 908–915 © 2014 The Authors Journal of Paediatrics and Child Health © 2014 Paediatrics and Child Health Division (Royal Australasian College of Physicians).
L Wang et al.
effects.16 The nicotine in tobacco is addictive and affects the appetite of tobacco users17 with strong nutritional implications. Studies have found that maternal behavior that impacts the intrauterine environment is essential to the development of the offspring18,19 and that maternal smoking and SHS potentially impact the weight status of children.20,21 Thus, there is a need to study maternal smoking behavior and/or exposure to SHS because of the implications for the child development, including risk of childhood and adolescent obesity.22,23 Emerging evidence suggests that prenatal environments may have long-term effects on health outcomes of children,24 which imply a possible association between prenatal factors and obesity among offspring. Previous studies have explored association between prenatal maternal smoking and weight of offspring using body mass index (BMI) measurements of groups ranging from newborn to 33 years of age.25–31 The magnitude of the risk and the dose-response for these relationships vary by age group and sex.31 Children exposed to cigarette smoking in utero are more likely to have a lower birthweight than those without exposure22 and are conversely at an increased risk of being obese.23,30 It has been found that SHS is independently associated with reduced anthropometric measurements of the newborn and increased rate of low birthweight infants.32 A cross-sectional study in the United States examining the impact of maternal and paternal smoking during pregnancy and the risk of overweight and obesity observed the strongest association with adult overweight and obesity when both parents smoked, although BMI was based on self-reported height and weight.21 A prospective study in the Netherlands examined only the independent effect of maternal smoking during pregnancy and ‘smoking in the parental house’ on childhood overweight without assessing the joint effect of smoking and SHS exposure on the childhood overweight.20 Thus, little is known about the impact of prenatal maternal smoking or SHS exposure and the joint effects during the year before birth on adolescent obesity using longitudinal data and accurate anthropometric measurements. The Healthy People 2020 in the United States aims to reduce the proportion of adolescent obesity by 10% from 17.9% in 2005–2008,33 which generates the need to investigate sources of obesity among adolescents. It is the purpose of this study to evaluate the association between prenatal maternal smoking status, measured as either direct cigarette smoking or SHS exposure, during the year before birth and the risk of obesity in the offspring during adolescence. This study could inform smoking cessation among potential pregnant women and programmes to improve maternal and child health.
Method Study sample and data set This study utilised data from the National Institute of Child Health and Human Development (NICHD) Study of Early Child Care and Youth Development (SECCYD), a longitudinal study that followed a cohort of children at four continuous phases (1991–1994, 1995–1999, 2000–2004, 2005–2007), from birth through age 15.9 years. Participants were recruited from 24 hospitals in the vicinity of 10 data collection sites across the United States. A more complete description of data collection
Maternal smoking and adolescent obesity
procedures and protocols of the SECCYD can be found elsewhere.34 Families were enrolled in the study upon successful completion of data collection through the 1-month interview. Recruitment was accomplished during the first 11 months of 1991 and 1364 families were enrolled. We restricted the analyses to adolescents whose mothers completed the question regarding maternal smoking and SHS exposure during the year prior to birth and those who had weight and height data between ages 12.0 and 15.9 years. A total of 932 participants were included in the analytic sample. This study was approved by authors’ institutional review board.
Study variables Obese status of adolescents The outcome variable was adolescent obesity between ages 12.0 and 15.9 years, which is defined as a child whose BMI ≥ 95th percentile according to the CDC’s sex- and-age specific growth chart for the year 2000.35 The BMI of adolescents was calculated using measured height and weight obtained using standard protocols and procedures.30,34 Standardised procedures were used to measure height and weight during the interviews by the NICHD SECCYD staff. Height was measured with children standing without shoes, feet together and their backs against a calibrated 7-foot measuring stick. Weight was measured using a physician’s two-beam scale. Scales were calibrated monthly using certified calibration weights. Weight was measured with children in minimal clothing and recorded twice, each time to the nearest 0.25 pound (0.1 kg).
Prenatal maternal smoking The primary independent variables were mother’s smoking behavior or SHS exposure during the year immediately prior to the child’s birth (1990). These variables were assessed retrospectively at the time the child was 24 months old with questions relating to the ‘The Year before My Child Was Born’.36 Prenatal maternal smoking status was determined by asking mothers ‘Which sentence best describes your smoking pattern during the year before your child was born?’ The responses were (i) I did not smoke; (ii) I smoked but stopped before pregnancy began; (iii) I smoked, and stopped during the first 3 months of pregnancy; (iv) I smoked through the first 3 months but stopped before my child was born; and (v) I smoked throughout the year. Maternal SHS exposure was determined with the question, ‘During the year before your child was born, how frequently did you come in contact with family, friends or co-workers who smoked?’ and the responses were (i) never, (ii) sometimes, (iii) often and (iv) always. Because of the small sample size in some subgroups, mother’s smoking status during the 1 year before birth was combined into two categories: did not smoke (i) and smoked (ii–v). Because only two mothers smoked without exposure to SHS using the never versus sometimes/often/always classification, we re-categorised SHS exposure as ‘never/sometimes’ and ‘often/always’ to increase the sample size for this variable. Participants were categorised into four subgroups according to smoking and SHS exposure status: (i) neither (reference group, ‘the lowest risk group’); (ii)
Journal of Paediatrics and Child Health 50 (2014) 908–915 © 2014 The Authors Journal of Paediatrics and Child Health © 2014 Paediatrics and Child Health Division (Royal Australasian College of Physicians)
909
Maternal smoking and adolescent obesity
L Wang et al.
smoking only; (iii) SHS exposure only; and (iv) both smoking and SHS exposure (‘the highest risk group’).
stepwise fashion (Table 4). All data analyses were performed using SAS version 9.2 (SAS Institute Inc., Cary, NC, USA).
Covariates
Results
Maternal information collected at the time when the child was 24 months old included education (less than high school, high school and some college or above), poverty (above poverty line, at or below poverty line), employment (employed, not employed) and family structure (living together, not living together ). Additionally, self-reported maternal body weight in pounds and height in inches were assessed when the child was 15 years of age, and maternal BMI was calculated (weight (pounds)/height (inches2)* conversion factor of 703). Maternal obesity was defined as a BMI ≥ 30 kg/m2. Child information collected at 1 month after birth included breastfeeding (breastfeeding, not breast feeding), birthweight (kilograms) obtained from the medical chart, sex (male, female), race (White, non-White) and gestational age in weeks (calculated with the formula as integer (40*(birthdate − due date)/7). Screen viewing hours (i.e. hours per day watching TV, video, or DVD) were dichotomised into less than 2 h and 2 h or longer.
Table 1 compares participant characteristics in the analytic sample (n = 932) with the excluded sample (n = 432), showing that mothers in the excluded sample were more likely to have less than a high school degree (P < 0.0001) and to not be employed (P = 0.04). Except for sex (49.3% vs. 56.9% for male, P = 0.01), there was no significant differences between the analytic and excluded samples with respect to child factors. Table 2 shows the comparison of maternal and child factors according to the four subgroups of prenatal maternal smoking and SHS exposure. The proportions of mothers who had less than a high school degree (P < 0.0001), were at or below poverty level (P < 0.0001), were not employed (P = 0.006), were not living together with their partners (P < 0.0001), were not breastfeeding (P < 0.0001), and proportions of children who were non-Whites (P = 0.016) and had 2 h or longer screen
Statistical analysis Chi-square tests or Student’s t-tests were used to compare characteristics of participants in the analytic sample (n = 932) with those in the excluded sample (n = 432). Analysis of variance or chi-square tests were used to compare maternal and child factors according to the four subgroups of prenatal maternal smoking and SHS exposure. In this longitudinal study, adolescents’ weight and height were measured at multiple follow-up time points (grades 5, 6, 7, 8 and 15 years of age) from 12.0 to 15.9 years of age. Thus, adolescent obesity assessed at each time point should not be independent, but be correlated. The generalised estimating equation (GEE) was used to address this issue by using ‘working correlation matrix’ to account for the withinsubject correlation of response on dependent variables of different distributions.37 The missing child’s obesity data were assumed to be missing at random, a regular method in longitudinal studies.38 First-order auto-regression was used as the working correlation matrix for performing the GEE to account for correlations among repeated measurements in the pediatric obesity. GEE regression modeling approaches were employed to evaluate the independent and combined effects of prenatal smoking and SHS exposure on the development of adolescent obesity. In the first approach (entitled Model 1 in Table 3), we examined the univariate associations between prenatal maternal smoking and exposure to SHS and the development of adolescent obesity. In the second approach (Model 2), we examined the individual effect of prenatal smoking and SHS exposure in two separate multivariate regression models with adjustment for maternal factors (education, poverty status, family structure, breastfeeding status, employment, gestational age and maternal obesity) and child factors (sex, race and screen viewing hours per day). We added birthweight to Model 3 (the third approach) with adolescent obesity as the outcome to explore its role as a potential mediator. The joint effect of prenatal maternal smoking and maternal SHS exposure were evaluated in a similar 910
Table 1 Comparison of participant factors in the analytic sample versus the excluded sample (n = 1364) Factors
Maternal factors Education§ Less than high school High school Some college or above Poverty§ Below poverty line At or above poverty line Employment§ Not employed Employed Family structure§ Living together Not living together Child factors Birthweight (kg)¶ Sex§ Female Male Race§ White Non-White
Analytic sample† n = 932 (%)
Excluded sample‡ n = 432 (%)
73 (7.8) 189 (20.3) 670 (71.9)
66 (15.3) 98 (22.7) 267 (62.0)
108 (11.7) 813 (88.3)
35 (13.2) 231 (86.8)
281 (30.2) 650 (69.8)
101 (36.7) 174 (63.3)
797 (86.1) 129 (13.9)
235 (86.7) 36 (13.3)
P-value
