| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
From the Department of Social Medicine, University of Bristol, U.K. (D.A.L.); Child Development and Rehabilitation Services, Mater Children’s Hospital, Brisbane, Australia (M.J.O., W.B.); the School of Population Health, University of Queensland Medical School, Brisbane, Australia (A.A.M., G.M.W., J.M.N.); and the School of Social Science, University of Queensland, Brisbane, Australia (J.M.N.).
Address correspondence and reprint requests to Debbie A. Lawlor, PhD, Department of Social Medicine, University of Bristol, Canynge Hall, Whiteladies Rd., Bristol, BS8 2PR, U.K. E-mail: d.a.lawlor{at}bristol.ac.uk
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Methods: This study was a cross-sectional analysis of 3613 (1742 females) participants of an Australian birth cohort who were examined at age 14.
Results: Three hundred fifty-three (9.8%) of the participants had co-occurrence of three or four risk factors. Risk factors clustered in these adolescents with a greater number of participants than would be predicted by assumptions of independence having no risk factors and three or four risk factors. The extent of clustering tended to be greater in those from lower-income families and among those with lower cognitive function. The age-adjusted ratio of observed to expected co-occurrence of three or four risk factors was 2.70 (95% confidence interval [CI], 1.80–4.06) among those from low-income families and 1.70 (95% CI, 1.34–2.16) among those from more affluent families. The ratio among those with low Raven’s scores (nonverbal reasoning) was 2.36 (95% CI, 1.69–3.30) and among those with higher scores was 1.51 (95% CI, 1.19–1.92); similar results for the WRAT 3 score (reading ability) were 2.69 (95% CI, 1.85–3.94) and 1.68 (95% CI, 1.34–2.11). Clustering did not differ by sex.
Conclusion: Among adolescents, coronary heart disease risk factors cluster, and there is some evidence that this clustering is greater among those from families with low income and those who have lower cognitive function.
Key Words: cardiovascular risk • cognitive function • socioeconomic position • epidemiology • adolescence
Abbreviations: BMI = body mass index; CHD = coronary heart disease; CI = confidence interval; MUSP = Mater University Study of Pregnancy and its outcome.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
There is increasing evidence that the pathological processes involved in the development of atherosclerosis are established in childhood (4–6) and that risk factors from across the life course are important in the development of cardiovascular disease (7). Many of the health risky behaviors begin in adolescence and continue over the life course. For example, adolescence is a particularly sensitive period with respect to learning about and adopting smoking behaviors, which then persist into adulthood (8,9). The recent emergence of type 2 diabetes among obese adolescents also highlights the importance of this period of the life course with respect to establishing other CHD risk factors (10). Although a number of studies have found components of the metabolic syndrome cluster in children and adolescents (11–14) and have focused on pathological causes of this syndrome, few have examined the role of social and cognitive factors in risk factor clustering or examined risk factors beyond those that form the metabolic syndrome. One exception to this is the work of Jessor and colleagues, which suggests that a single underlying factor could account for the modest correlations between a range of health-enhancing behavior, including seatbelt use, adequate hours of sleep, healthy diet, adequate exercise, and regular toothbrushing (15,16).
Because individual behavioral risk factors such as smoking and physical inactivity are strongly socially patterned and also influenced by cognitive ability (17–19), socioeconomic position and cognitive function may be particularly important in the clustering of behavioral risk factors or physiological risk factors that are determined by behavior. A study of adults in The Netherlands found that co-occurrence (the existence in an individual of a large number of risk factors), but not clustering (a greater-than-expected number of individuals with either no risk factors or a large number of risk factors), of risk factors was associated with educational attainment, (20) and recently, similar results were found for occupational social class in a group of older British women (21). However, the role of socioeconomic position in risk factor clustering may be more important in adolescence when experimentation with, and adoption of, risky behaviors begins. To our knowledge, no previous study has assessed whether there is differential clustering of risk factors by cognitive ability, but because cognitive ability is an important predictor of behavioral risk factors in adolescence (19,22), it is plausible that this will also influence the extent of clustering.
The aims of this study are: 1) to examine the extent of clustering of four CHD risk factors—smoking, high levels of television watching (used as a proxy for physical inactivity), overweight, and high blood pressure (both of which are influenced by dietary and physical activity behaviors)—among 14-year-old Australians; 2) to determine whether clustering differs by gender, according to socioeconomic position and by cognitive function; and 3) to assess the association of socioeconomic position and cognitive function with having three or four of these risk factors compared with having fewer.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
The mothers and children have been followed up prospectively with maternal questionnaires being administered when their children were 6 months, 5 years, and 14 years. In addition, at 5 and 14 years, detailed physical, cognitive, and developmental examinations of the children were undertaken, and at 14 years, the children completed health, welfare, and lifestyles questionnaires. Of the original 7223 participants, 5172 (72%) of the children participated in the 14-year assessment. Children who did not participate were more likely to be from families with low income at birth, to have mothers who smoked throughout their pregnancy, and mothers and fathers with lower educational attainment (24).
Measurements
At age 14, the children were asked if they smoked cigarettes and were given the response options: often, sometimes, never, or rarely. They were further asked "in the last week, how often have you smoked?" (everyday, every few days, once/so, not at all) and "how many cigarettes have you smoked in the previous week?" (nil, 1–9, 10–19, 20–29, 30–49, 
50 cigarettes in the previous week). Smoking at age 14 was categorized as never (answered never/rarely to first question, not at all to second, and nil to third) or smoker (answered sometimes or often to first question, at least once/so to second, and at least 1–9 cigarettes per week to the third question). This classification resulted in 96 (1.9%) of 5170 with smoking data at age 14 who could not be classified because of contradictions in their answers, i.e., suggesting in one or more question that they smoked but in at least one question that they did not. For the main analyses presented here, these children were allocated to the smoking category. Two sensitivity analyses were undertaken in which these children were allocated to the nonsmoking category and were excluded. The results of these sensitivity analyses did not differ from those presented here. The children were also asked to report how many hours of television they watched per day on average Monday to Friday and also how many hours they watched on average per day Saturday and Sunday (for both, six category responses were provided: never, less than 1, 1 to <3, 3 to <5, 5 to <7, 7 or more hours). There was a high level of agreement between hours watched during the week and during the weekend, and for this study, high television watching was defined as watching 5 or more hours per day Monday to Friday. This measure is used as a proxy indicator of physical inactivity and consumption of high-fat/high-sugar convenience foods among the children (25,26).
The average of two measures of the child’s weight, lightly clothed with a scale accurate to 0.2 kg, was used in all analyses. Height was measured using a portable stadiometer, and triceps skinfold thickness was measured using standard skin calipers. Overweight was defined according to standard definitions derived from international surveys by Cole et al. (27). Thus, in this study, the 14 year olds were defined as overweight if their body mass index exceeded 22.62 kg/m2 for boys and 23.34 kg/m2 for girls, which are equivalent to exceeding 25 kg/m2 in adulthood (27). Two measures of blood pressure were taken 5 minutes apart using a digital sphygmomanometer with the child seated and at rest. The mean of these two measures was used in all analyses. Because there is no standard definition for high blood pressure in childhood and because systolic and diastolic blood pressure are highly correlated, mean arterial blood pressure was calculated as: diastolic blood pressure + one third (systolic blood pressure – diastolic blood pressure), and those above the sex-specific 75th percentiles of mean arterial pressure were defined as having high blood pressure. This gave cutoffs of 86.0 mm Hg for females and 87.0 mm Hg for males.
Raven’s standard progressive matrices (28) and the Wide Range Achievements Test version 3 (WRAT3) (29) were used to assess cognitive function. The Raven’s standard progressive matrices (Raven’s SMP) is a test of nonverbal reasoning ability that has been widely used for psychological assessment in clinical and educational contexts for research and for personnel selection (28). The Raven’s SMP scores were age-standardized in 6-monthly intervals. The WRAT3 is an age-normed reference test that assesses reading and word decoding skills (29). It is reliable, predictive of future educational attainment, and has been widely used in research (30). Mothers reported the gross family income when the child was aged 14 in seven categories with each increasing by approximately AU $5000 (23). In the analyses in which the extent of clustering was compared between strata of family income, low income was defined as AU 
$25,999, as in previous reports from the MUSP (23).
Statistical Analyses
The age-adjusted (in days) prevalence of each risk factor was tabulated by sex, income, and cognitive function. We calculated expected frequencies of co-occurrence of risk factors using these age-adjusted prevalences by combining probabilities assuming a binomial distribution and independence between each risk factor. Observed-to-expected ratios for each combination of risk factors were calculated for the whole cohort. To examine whether the extent of clustering (greater co-occurrence than would be predicted if the four risk factors were independent of each other) varied by sex, income, and cognitive function, we dichotomized the income and cognitive function variables and compared the observed-to-expected ratios for zero to four risk factors in those of low income and low cognitive function with those with higher income/cognitive function. To compare the extent of clustering in different groups by cognitive function and income, it was necessary to stratify these variables. Both cognitive development scores were categorized as low (at or below the 25th percentile for the whole cohort) or above the 25th percentile, like in previous studies (28,29) Low family income when the child was aged 14 was defined as AU $25,999, as in previous reports from the MUSP (23), and participants in this category were compared with those with greater family income. In these clustering analyses, expected frequencies were those predicted given the prevalence of risk factors within each sex, income, or cognitive function group. Clustering is indicated when observed:expected ratios are high for having no risk factors, low for having only a single risk factors, and high for having co-occurrence of three or four risk factors (that is, if clustering is present, individuals are more likely to have either none or many risk factors compared with expected). To further assess clustering, a negative binomial regression model was fitted to the counts of risk factors, and a likelihood ratio test was used to assess whether the overdispersion parameter differed from zero.
In addition to this clustering analysis, we examined whether income and cognitive function were independently associated with simple co-occurrence of risk factors. Here, we used multiple logistic regression to assess the associations of low income and cognitive function with the co-occurrence of three or four risk factors and, in addition, assessed the association of both exposures as continuous variables with the co-occurrence of three or four risk factors. All analyses were undertaken using Stata version 8.0 (Stata Inc., TX).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Table 1 shows the age-adjusted prevalence of each risk factor in the whole analysis cohort and by sex, family income, and cognitive function. Males were more likely than females to watch 5 or more hours of television per day, but other risk factors did not vary by sex. Those from low-income families and with low cognitive function (both nonverbal reasoning and reading ability) were more likely to smoke, watch 5 or more hours of television per day, and be overweight, although the difference in television watching did not reach conventional levels of significance with reading ability (WRAT 3). The prevalence of high blood pressure did not vary by sex, family income, or cognitive function.
|
In total, 353 (9.8%) of the 3613 participants had co-occurrence of three or four risk factors. Figures 1, 2, and 3 show the ratios of observed-to-expected prevalence of co-occurrence of risk factors for the whole cohort and by sex, family income, and cognitive function. There was clustering within the whole cohort with the ratio of observed-to-expected prevalence of having none and three or four risk factors being greater than one and the observed-to-expected ratios of having just one or two risk factors being less than one. The overdispersion parameter for the whole cohort differed significantly from zero (p < .001). In stratified analyses by gender, family income, and cognitive function, clustering occurred in all subgroups (p < .03 for overdispersion in all models). That is to say, there was statistical evidence that these risk factors clustered in both sexes, in those from low- and higher-income families and in those with low and higher cognitive function scores. Examination of Figures 1 to 3 and of the point estimates in the stratified analyses provides some suggestion that the extent of clustering was greater in those from lower-income families and with lower cognitive function. The age-adjusted ratio of observed-to-expected co-occurrence of three or four risk factors was 2.70 (95% confidence interval [CI], 1.80–4.06) among those from low-income families and 1.70 (95% CI, 1.34–2.16) among those from more affluent families. The ratio among those with low Raven’s scores (nonverbal reasoning) was 2.36 (95% CI, 1.69–3.30) and among those with higher scores was 1.51 (95% CI, 1.19–1.92); similar results for the WRAT 3 score (reading ability) were 2.69 (95% CI, 1.85–3.94) and 1.68 (95% CI, 1.34–2.11). However, there was no strong statistical evidence that the extent of clustering was different between those in low- compared with high-income families (p = .2) or between those with low compared with high cognitive function (p value for Raven’s score = 0.5 and for WRAT 3 = 0.6). Stratified analyses did not suggest that clustering differed by sex.
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Study Limitations
Seventy-two percent of the original birth cohort participated when they were aged 14, but in this analysis, only 70% of these respondents had complete data on all relevant variables. Nonparticipants at age 14 were likely to be from more adverse socioeconomic backgrounds at birth (31), but there were no important differences between those participants at age 14 with complete data for analyses and those without these complete data. Our results would only be biased if the associations we have found were nonexistent or in the opposite direction among those who did not participate or who had incomplete data, which seems unlikely. We do not have data on lipid levels or indicators of glucose intolerance or insulin resistance, and therefore could not include these risk factors in our analyses. We do not have detailed assessment of physical activity or physical fitness and therefore used hours of television watching, which has been shown to be associated with low levels of physical activity and high consumption of poor-quality, high-sugar and -fat content foods (32,33) Despite several studies showing strong associations between levels of physical activity and poor-quality diet with television watching, this variable may also reflect a range of other factors, including family characteristics such as parental engagement with their children. Thus, similar analyses as those presented here but in studies with direct measurements of physical activity and diet would be useful. Although our stratified analyses suggested that there was a greater extent of clustering in those from low-income families and with low cognitive function, we had limited statistical power to detect a true difference in clustering between these groups. Replication of our findings in larger studies would be valuable.
Implications
Two previous studies in adults found that CHD risk factor clustering did not differ between those with low and high educational attainment (34) or between those from lower and higher occupational social class (35). Our finding that clustering in adolescence tends to be greater among those from lower-income families and with lower cognitive function may be because adolescence is a particularly sensitive period for experimentation with risky behaviors. Our results suggest that not only are poor socioeconomic position and lower cognitive function associated with increased likelihood of CHD risk factors, as found in previous studies (19,36,37), but that in adolescence, poorer socioeconomic position and lower cognitive function may be associated with greater clustering of these risk factors. The association of postmortem atherosclerosis in adolescents with established CHD risk factors such as smoking, obesity, and higher blood pressure suggests that the existence of these risk factors in adolescence is associated with established arterial disease, which is likely to be irreversible (6). Previous studies in adolescence have also found clustering of health-related behaviors in adolescence with the suggestion of a single underlying causal factor (15,16) In that work, the clustering of health-enhancing behaviors such as healthy diet, adequate sleep and exercise, wearing of seatbelts, and dental hygiene was predicted by greater psychosocial conventionality, whereas risky behaviors such as problem alcohol drinking, smoking, and illicit drug use clustered separately and were predicted by unconventional traits (16). That work was conducted in the United States nearly 2 decades ago and its relevance to contemporary adolescents is unclear. However, of interest, the same team of investigators in a recent publication also found two latent variables underlying adolescent behaviors (38). Sexual intercourse, alcohol abuse, illicit drug use, delinquency, and smoking strongly loaded on one latent variable (labeled "problem behavior" by the investigators), whereas unhealthy behavior, sedentary behavior, and poor health all loaded together on a separate latent variable (labeled "health-related behaviors"). The authors concluded that adolescent cigarette smoking relates strongly and directly to problem behaviors and only indirectly, if at all, to health-compromising behaviors (38). However, it should be noted that latent variable analyses such as these are to some extent driven by the variables included in the model, and the results in this study suggest that among contemporary U.S. adolescents cigarette smoking is more strongly correlated with sexual intercourse, alcohol abuse, illicit drug use, and delinquency than it is with other health risky behaviors such as poor diet and lack of physical activity. These findings do not show that cigarette smoking is not correlated with these other health-related behaviors.
Our results suggest that health promotion initiatives in adolescents need to be aimed at improving socioeconomic circumstances and cognitive function, which influence the co-occurrence of risk factors and may also affect the extent of their clustering in individuals. Although political initiatives could result in improvements in the socioeconomic circumstances of those from the lowest socioeconomic groups, there is much debate about whether cognitive function is primarily determined by genetic factors (nature) or environmental factors (nurture) that might be modifiable. Twin and adoption studies suggest heritability estimates of approximately 0.50 for cognitive function with modest common environmental influences (39,40), whereas, elsewhere, the relative contributions of genetic and environmental contributions to childhood intelligence appeared to vary by family income (41). Thus, among children brought up in impoverished families, 60% of the variance in intelligence scores at age 7 years was accounted for by shared environmental factors with very little contribution from genetic factors. By contrast, among children from affluent families, most of the variation in intelligence was accounted by genetic factors. Further support for the importance of environmental factors in determining cognitive function in early life can be found in trials of early learning and school-readiness programs. In two recent systematic reviews of such interventions (42,43), one of which focused on randomized trials (42), the conclusion of both was that these programs had important effects on reading, arithmetic ability, and general intelligence that extended to secondary school ages (42,43) Finally, the observation of secular increases in cognitive function across a range of populations is strong evidence for an important environmental effect. In these studies, the increases—widely referred to as the Flynn Effect—have occurred far too quickly for them to be explained purely by changes in the gene pool (43). A similar argument is widely cited, and accepted, for the role of environmental factors in the so-called obesity epidemic. Thus, although genetic factors may be important determinants of individual levels of cognitive function, there is also considerable evidence for the importance of modifiable environmental factors. Long-term follow up of trials of early learning programs, which have found improvements in childhood cognitive function, to examine their effects on CHD behavioral risk factors would provide valuable insights into the causal associations between cognitive function and these risk factors. They would also provide insights into the potential of improving CHD risk through interventions aimed at improving cognition.
Our results also suggest that interventions to improve CHD risk in adolescents need to be aimed at tackling a range of risk factors rather than focusing on just one risk factor such as smoking or physical inactivity. Furthermore, this work suggests that socioeconomic position and cognitive function are characteristics that might be used to identify groups that are at particularly high risk of future health problems.
The authors are grateful to all participants in the study. Greg Shuttlewood, University of Queensland, helped with data management for the study.
|
NOTES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
D.A.L., M.O., and A.A.M. developed the study aim and design. W.B., J.N., M.O., and G.W. set up and are responsible for the conceptual development and continued management of the Mater University Study of Pregnancy and its outcomes. D.A.L. undertook the analysis and wrote the first draft of the paper. All authors contributed to the final version of the paper.
DOI:10.1097/01.psy.0000188576.54698.36
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  A. Alamian and G. Paradis Correlates of Multiple Chronic Disease Behavioral Risk Factors in Canadian Children and Adolescents Am. J. Epidemiol., November 15, 2009; 170(10): 1279 - 1289. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. Terre Cognitive Risks of Dyslipidemia American Journal of Lifestyle Medicine, July 1, 2009; 3(4): 284 - 286. [Abstract] [PDF]
|
|
|
|
 |
|
 A Vaananen, A Kouvonen, M Kivimaki, T Oksanen, M Elovainio, M Virtanen, J Pentti, and J Vahtera Workplace social capital and co-occurrence of lifestyle risk factors: the Finnish Public Sector Study Occup. Environ. Med., July 1, 2009; 66(7): 432 - 437. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C Power, K Atherton, and O Manor Co-occurrence of risk factors for cardiovascular disease by social class: 1958 British birth cohort J Epidemiol Community Health, December 1, 2008; 62(12): 1030 - 1035. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. C. Geerts, M. L. Bots, D. E. Grobbee, and C. S.P.M. Uiterwaal Parental Smoking and Vascular Damage in Young Adult Offspring: Is Early Life Exposure Critical?: The Atherosclerosis Risk in Young Adults Study Arterioscler Thromb Vasc Biol, December 1, 2008; 28(12): 2296 - 2302. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Kivimaki, D. A. Lawlor, G. D. Smith, A. Kouvonen, M. Virtanen, M. Elovainio, and J. Vahtera Socioeconomic Position, Co-Occurrence of Behavior-Related Risk Factors, and Coronary Heart Disease: the Finnish Public Sector Study Am J Public Health, May 1, 2007; 97(5): 874 - 879. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
