JOURNAL OF SPORTS SCIENCE & MEDICINE
Department of Public Health, University of Helsinki, Helsinki
(Online): 01 September 2003
This review is based on the following orginal publications, which will be referred to in the text as Studies I-IV:
I. Aarnio, M., Kujala, U.M. and Kaprio, J. (1997) Associations of health-related behaviors, school type and health status to physical activity patterns in 16 year old boys and girls. Scandinavian Journal of Social Medicine 25, 156-167.
II. Aarnio, M., Winter, T., Kujala, U.M. and Kaprio, J. (1997) Familial aggregation of leisure-time physical activity - a three generation study. International Journal of Sports Medicine 18, 549-556.
III. Aarnio, M., Winter, T., Peltonen, J., Kujala, U.M. and Kaprio, J. (2002) Stability of leisure-time physical activity during adolescence - a longitudinal study among 16-, 17- and 18-year-old Finnish youth. Scandinavian Journal of Medicine and Science in Sports 12, 179-185.
IV. Aarnio, M., Winter, T., Kujala, U.M. and Kaprio J. (2002) Associations of health-related behaviour, social relationships, and health status with persistent physical activity and inactivity: a study of Finnish adolescent twins. British Journal of Sports Medicine 36, 360-364.
Regular exercise has been shown to exert many positive effects on health. Sedentary behaviour often originates in childhood and many common adult chronic diseases are related to inactivity. Adolescent physical activity patterns and health habits are important subjects to study because of the known associations of physical activity with other health habits and the evidence that these associations track into adulthood.
The data for this study were gathered as a part of FinnTwin16, a longitudinal study of five consecutive birth cohorts of Finnish twins, their siblings and parents. The study material was collected by identifying twins born in 1975-1979. Questionnaires concerning leisure-time physical activity, health-related behaviours, social relationship and health status were sent to twins on their 16th and 17th birthdays, and six months after their 18th. The maximal cohort size was 4906 boys and girls, and the response rate 75.8% to 81.7%.
The results of this study reveal that persistently active adolescents smoked less than inactive ones, and usually had better health and nutritional habits, such as use of spreads and regular breakfast eating, and better self-estimated health. They attended high schools rather than vocational schools and tended to have better academic achievement. Participating in organised sport, in many different types of sport, or in power sports and ball games were also associated with persistent physical activity. Parents' and grandparents' physical activity were not associated with adolescent physical activity except in the case of very active mothers and daughters, but a co-twin's physical activity was associated.
There was a gender difference in physical activity patterns: boys were more active than girls. No gender difference was found in health related-behaviours, except that girls reported more psychosomatic symptoms such as tension, in the low physical activity categories than boys.
The known health benefits of physical activity and risk of declining activity during adolescence makes young people at this stage of life an important target group for physical activity promotion programmes. Based on this study, physically active adolescents seem to progress to a healthier and more educated life. Adolescents approaching the completion of their compulsory education and thus the end of their systematic physical education can be considered a particular risk group for inactivity.
KEY WORDS: Leisure activities, exercise, health status, food habits, smoking, life style, family health, family relations, adolescence, adolescent behavior, twins, cohort studies.
Regular exercise has been shown to have a number of positive effects on health. Among adults, higher levels of physical activity have been associated with reduced incidences of cardiovascular disease (Kohl, 2001), hypertension (Fagard, 2001), low back pain and osteoporosis (Vuori, 2001), depression (Dunn et al., 2001), and certain types of cancer (Thune and Furberg, 2001). Numerous risk factors for coronary artery disease, hypertension, non-insulin-dependent diabetes and osteoporosis appear to develop in childhood and youth (Williams et al., 1981). Many of the risk factors for coronary heart disease are related to reduced physical activity and sedentary behaviour, both of which commonly originate during childhood (Sallis et al., 1992).
Healthy children are naturally active, but then tend towards inactivity as they grow older (Raitakari et al., 1996). During recent years society in general has become more inactive; for example, children no longer walk to school but take the bus. Physical activity classes have decreased in Finnish schools as a result of economic cutbacks in the 90s. However, whereas habitual physical activity seems currently to be on the decline, participation in organised sport and the amount of time devoted to physical activity increased during the 90s.
The relatively few longitudinal studies of physical activity have found wide variations in the stability of physical activity pattern correlations. According to Kelder et al. in a seven-year follow-up, students identified at baseline with high physical activity tended to remain at this level, as did those with initially low physical activity (Kelder et al., 1994). In a Finnish study of adolescents, however, tracking correlations of physical activity were statistically significant but rather low (Telama et al., 1996).
Health related behaviour and social relationships also seem to be associated with physical activity. Kelder et al. (1994) demonstrated in their study some evidence of consolidation and tracking of physical activity, smoking behaviour and food preference. Sallis et al. (2000) reviewed 108 studies of physical activity among children (aged 3-12) and adolescents (aged 13-18). Variables that were consistently associated with adolescent physical activity were sex (male), ethnicity (white), age (inverse), perceived activity competence, depression (inverse), previous physical activity, community sports, sensation seeking, sedentary after school and on weekends (inverse), encouragement from parents and others, sibling physical activity, and opportunity to exercise.
What are the factors involved in persisting with physical activity or giving up? According to Telama et al. (1994) declining physical activity is associated with the phase of puberty, and pubertual changes might be associated with reduced physical activity (Yang, 1997). Also, the gender difference has been pronounced in most studies, boys being more active than girls (Sallis, 1993; Hickman et al., 2000; Hämäläinen et al., 2000). As the association of physical activity with other health habits among children and adolescents is known, and there is some evidence that these carry into adulthood, adolescent physical activity patterns, health related behaviour and possible gender differences offer a productive and important field of study.
Determining and measuring physical activity
According to Telama and Laakso (1983) it is important to distinguish between habitual physical activity and that which occurs mainly in leisure time, e.g. in sports clubs. Measurement of physical activity can include frequency, intensity or duration of physical activity. Caspersen et al. (1985) define physical activity as "any bodily movement produced by skeletal muscle that results in energy expenditure". This work concentrates on leisure-time physical activity.
Sirard and Pate (2001) reviewed 59 articles concerned with validating physical activity measurement methods in children and adolescents. They categorised the measures used as primary, secondary and subjective. In this review, direct observation, doubly labelled water and indirect calorimetry were considered the primary standards for assessing physical activity. Heart rate monitors, pedometers and accelerometers were considered as secondary measures, because they provide an objective assessment of physical activity. Finally, surveys, self-report questionnaires, interviews, proxy-reports and diaries were considered as subjective techniques (Sirard and Pate, 2001). According to this review direct observation offers the most practical and appropriate measure of physical activity and patterns of activity. Correlations of direct observation with heart rate or oxygen consumption were 0.61 to 0.91. All the primary measures are reliable, but rather expensive and not appropriate for the large scale of epidemiological studies. Special equipment may also be needed. Secondary measurement devices such as heart rate monitors and motion sensors are also useful in physical activity studies.
In epidemiological studies, interviews and questionnaires are the most useful and cost efficient means of measuring physical activity. For example, in a Finnish national survey, the Adolescent Health and Lifestyle Survey (AHLS), a questionnaire for physical activity was used. The frequency of physical activity was measured by asking the amount of time spent in leisure time physical activity and participation in organised sports. The intensity of physical activity was measured by asking about breathing intensity and sweating during physical activity, using five alternatives (Hämäläinen et al., 2000).
Distribution of physical activity
Trends in physical activity of 12-, 14-, 16- and 18-year-old Finns have been studied in a nationally representative biannual postal survey, the Adolescent Health and Lifestyle Survey, from 1977 to 1999. The questionnaire is sent to Finnish adolescents aged 12-18. The number in the total cohort in 1999 was 70351, and the response rate has ranged from 76% to 88%. Physical activity is measured by frequency, intensity and participation in organised sport (Rimpelä et al., 1999). According to Hämäläinen et al. (2000) the 1999 AHLS survey shows that activity rate (especially sports club activities) decreased with age, while perceived intensity of activity increased. Participation in sports club activities, perceived intensity of activity and the proportion of those reporting very frequent physical activity increased during the study period. Physical activity outside sports clubs remained more common than participation in sport club activities (Hämäläinen et al., 2000).
According to a WHO adolescent survey (Kannas and Tynjälä, 1998) the proportion of very active boys increased in Finland between 1986 and 1998. Both time and participation in sports clubs have increased. In general, leisure time physical activity has expanded among adolescents since 1980 (Kannas and Tynjälä, 1998; Nupponen and Telama, 1998). The proportion of the very active has increased among boys of all age groups.
In a pan-European WHO study, Health and Health Behaviour of Young People, adolescents aged 11, 13 and 15 years were asked how often and how many hours a week they took part in vigorous intensity activity outside school hours. Vigorous physical activity was defined as equivalent to at least slow jogging which might be expected to leave the participant feeling out of breath and sweaty (Hickman et al., 2000). The total number interviewed was 123 227 in 29 European countries, plus USA, Canada and the Russian Federation. The number in the Finnish cohort was 4864. Overall, students who reported exercising at least twice a week in 1997/1998 were most prevalent in Northern Ireland, Austria, Scotland and Estonia, and in the 15-year-old group also in Germany and the Czech Republic. The most inactive students were from Greenland, Latvia, Lithuania and Hungary. The proportion of students who reported exercising at least twice a week in Finland ranged from 79% to 86% among boys and 64% to 74% among girls across the age groups. The students who reported exercising two hours a week or more were most often from Austria, Denmark, Switzerland and Germany, while those countries with the lowest proportion of adolescents exercising at this level were Latvia, Portugal and the Russian Federation. The proportion of Finnish adolescents in this category ranged from 70% to 76% among boys and 56% to 60% among girls. Finland was also among the most active countries (Hickman et al., 2000).
A gender difference was apparent in both surveys, boys being more active than girls. In the ALHS survey 65% of boys and 55% of girls were participating in physical activity four times a week or more at the age of 12. Perception of being very active declined with age between 12 and 18 years: among boys from 26% to 12% and among girls from 13% to 5%. At the age of 12, 58% of boys and 43% of girls were participating in intense physical activity (sweating and breathing intensely). Very intense physical activity increased from 12 to 18 years: among boys from13% to 43% and among girls from 5% to 17% (Hämäläinen et al., 2000).
According to the WHO survey 82% of boys aged 11-15 in 1998 participated in physical activity at least twice a week, and 52% four times a week; the corresponding proportions in 1996 were 67% and 34%. Among girls in 1998, 68% took physical activity twice and 33% four times a week. In 1996 the corresponding figures were 53% and 18% (Kannas and Tynjälä, 1998). In 1998 6% of boys and 11% of girls were inactive. The proportion of inactive youngsters (physical activity less than once a week) decreased between 1996 and1998 among boys, while among girls it varied according to age group (increased among 11-year-olds, stayed the same among 13-year-olds and decreased among 15-year-olds) (Kannas and Tynjälä, 1998).
Based on a WHO European survey, Health Behaviour in School Aged Children (Hickman et al., 2000), in most of the countries vigorous activities are more common among boys than girls and decline with age, especially among girls. The gender differences are pronounced in most countries. For example, the proportion of girls exercising vigorously is approximately half that of boys among 15-year-olds in Greenland, Lithuania and Greece (Hickman et al., 2000). Eleven-, 13- and 15-year-old boys and girls are most likely to exercise regularly in Northern Ireland and Austria, while the lowest proportion doing regular physical activity is in Greenland (Hickman et al., 2000).
Stability and changes of physical activity
Other Nordic studies have also tended to show that physical activity declines sharply during the early years of adulthood (Andersen and Schelin, 1994; Raitakari et al., 1996). In the 1970s a longitudinal Swedish study reported a significant decline in numbers of the physically active, as well as in the amount of physical activity, between the ages of 15 and 20 (Engström, 1980).
Sallis (1993) reviewed nine studies employing standardised self-reports or objective measures of physical activity. These studies reveal a consistent decline in physical activity over the school-age years, with the mean level of physical activity decreasing about 2.7 % per year among males and 7.4 % per year among females. Kelder and colleagues' (1994) studies of American adolescents from the 6th to 12th grades found that high activity levels at baseline tended to remain high as the students aged, and that low activity levels at baseline were also persistent. Most of the studies in Table 1 report low to moderate tracking of physical activity.
There have been a few large longitudinal studies in Europe. Malina (1996) compared three major European studies tracking physical activity among Finnish (Raitakari et al., 1994) and Dutch adolescents (van Mechelen and Kemper, 1995), and Belgian boys (Vanreusel et al., 1993). Correlations (Pearson or Rank order) under 0.30 were considered low, those between 0.30 and 0.60 moderate, and those over 0.60 high (Malina, 1996). Correlations for estimates of physical activity over three-year periods during adolescence were fairly uniform in the three studies, at 0.33 to 0.44, the one exception being Dutch girls with a correlation of 0.58. In the Dutch study, the three-year correlation in boys for energy expenditure in organised sports was higher than that for physical activity - 0.53 and 0.44 respectively - while the corresponding correlations were essentially the same in girls, at 0.58 and 0.59. Correlations for physical activity over a longer period were much lower for Finnish youngsters (0.18 and 0.17) over the six years from 12 to 18, while the 0.37 correlation in Belgian boys between 13 and 18 did not differ from the 0.35 found between 13 and 16 years. The results of these three studies are generally consistent in indicating low to moderate tracking of physical activity during adolescence (Malina, 1996).
It is clear that with increasing follow-up time the tracking correlation
drops; in Table 1 those studies with
shorter follow-up time have higher tracking correlations. Such correlations
are therefore not comparable without taking the follow-up time into consideration.
Adolescent physical activity is heavily promoted in Finland by individual organisations within the Finnish Sports Federation (Nuori Suomi /Young Finland), which receives government support. According to Nuori Suomi most adolescents are independently active outside home, either alone or with friends. Second comes membership of a sports club; in Finland 350 000 3-18-year-old children and adolescents belong to a sports club. The most popular sports, depending on age and the type of questionnaire used, are jogging, walking, swimming, football and floorball (hockey played indoors) (Nuori Suomi). There are gender disparities in preferences for sports. Boys tend to participate more in football, ice-hockey, floorball, basketball, weight lifting, and track and field, while girls prefer jogging, walking, riding, gymnastics, slalom and aerobics (Nupponen and Telama 1998).
Playing organised sport has an important influence on physical activity later in life. Participation in competition sports (Telama and Yang 1997) and membership of a sports club (Barnekow-Bergkvist et al. 1998) are predictors of later physical activity, but the best predictor is the student's school grade for physical education, and participation in organised sports (Telama et al. 1994). According to Telama et al. (1996) the highest tracking correlation was for frequency of participation in sports clubs, with the correlation varying from 0.40 to 0.78 among boys and from 0.28 to 0.64 among girls. This may be because the stability of non-organised physical activity appears to be lower than that of organised sports, as it is more difficult to estimate and remember non-organised activities (Telama et al.,1996).
Correlations of health-related behaviours, social relationships and
health status with physical activity
This review focuses on those health-related variables shown to be related to physical activity among adults and adolescents, and that are predictors of morbidity and mortality, in particular from cardiovascular diseases.
Associations of smoking with physical activity
Association of other health-related behaviours and health status with
The Amsterdam Growth and Health Longitudinal study found that although adolescent physical activity was not related to adult cardiovascular health status, physical activity was directly associated with serum HDL-cholesterol levels (van Mechelen et al. 1999) and lower consumption of saturated fatty acids (Raitakari et al. 1994). According to Yang et al. (1999), early physical activity and current social and health-related behaviours were significantly related to the level of adult physical activity.
Associations of social relationships with physical activity
Longitudinal studies show that certain social and environmental factors may predict consistent physical activity. These include higher school grades and participation in organised sports (Telama 1994), and playing sport for school (Dovey et al. 1998), as well as social relationships (Yang et al. 1999), the adolescent's local environment (Telama et al. 1994), and very good self-assessed health (Dovey et al. 1998). The studies of correlations between smoking, other health-related behaviours and physical activity are seen in Table 2.
Familial aggregation of physical activity
Results of studies of parental impact on adolescents' physical activity patterns vary tremendously. Findings from comparing the association of fathers' or mothers' activity with adolescent physical activity are also contradictory. According to Rossow and Rise (1994), fathers' physical activity was positively associated with their adolescent's physical activity, but mothers' physical activity was not. Yang et al. (1996) reported similar results from their 12-year follow-up study, in which fathers' physical activity was associated with their adolescent's physical activity in the same year and was a significant predictor of boys' and girls' physical activity 12 years later and for boys even longer. Lau et al. (1990) reported no significant association between the physical activities of mothers and their adolescents. The studies are seen in Table 3.
Gottlieb and Chen (1985) found that parental exercise had a stronger influence on the frequency of exercise among girls than boys, but according to a study of Finnish school children (Telama and Laakso 1983) both parents' sports activities were correlated with boys' sports activities, but not with girls'.
In general, there seems to be a significant association with family members' and friends' physical activity. The WHO cross-national study on health behaviours among 11-, 13- and 15-year-old adolescents in 10 European countries indicates that when three or more significant persons (family members or best friends) take part in physical activity, 84% of boys and 71% of girls are involved in sport twice a week or more (Anderssen and Wold 1992). When none of these significant others is involved in physical activity, only 52% of boys and 30% of girls report being active in sport.
The present review does not take into consideration the genetic influence in familial aggregation of physical activity or the causality of friends' leisure time physical activity, because it is too large an issue and needs further study. Family members share both genetic background and environment, and friends only the environment. Physical activity is very much a social activity in adolescence. Young people tend to perform these activities together, and it is consistently shown that physically active adolescents have friends who are also active (Anderssen and Wold 1992).
The aims of this study were:
The FinnTwin16 study material was collected by identifying twins born in 1975-1979 and their parents from the Central Population Registry of Finland. The baseline assessment was made within two months of the twins' 16th birthdays. It included a survey of health related behaviour and attitudes, a symptom check-list, and relationships with parents, peers and the co-twin.
To 16-year-old twins born in 1975-1979 the questionnaire was mailed in 1991 through to1995. Parents were also sent a questionnaire with the same types of questions. It also included items concerning the maternal and paternal grandparents' related behaviour, leisure time physical activities and socio-economic status.
Two further questionnaires were sent to all twins who replied at the age of 16. One was sent a month after their 17th birthday, and another about six months after their 18th birthday (mean response age 18.5 years). The fourth wave of assessment of the twins is ongoing in 2000-2002.
This thesis is based on four articles (I, II, III, IV), and the material were based on the FinnTwin 16 cohort study (Figure 1a, 1b). The number of twins varies between the articles because the total follow-up data are collected in five-year periods, and in the first articles the available data are based on the material collected over a three-year period. The first article (I) is based on the questionnaire sent to twins on their 16th birthday in 1991-1993. The article explores the association of physical activity at the age of 16 with health-related behaviours, social relationships and health status. The kind of school attended after the end of compulsory education is an extra item based on the 17-year questionnaire, because compulsory universal schooling ends at the age of 16. During this three- year period, 1858 families of twins (boys or girls) were contacted and the total number of individual twins who answered was 3254 (girls n=1697, boys n=1557). The response rate was 91% for girls and 85% for boys (I). The second article (II) concentrates on familial aggregation of physical activity. Information about parents' and grandparents' physical activity was based on the questionnaire sent to families and twins at the age of 16. The number of families with both parents was 1667 and the response rate among parents was 79% (II). Leisure-time physical activity was assessed using different questions for adolescents, parents and grandparents. Whole families consisted of a pair of adolescent twins, father and mother and four grandparents.
The third (III) article studying the stability of physical activity and types of sport was based on a questionnaire sent to twins at their 16th, 17th and 18.5th birthday. The stability of physical activity was based on all three questionnaires, and the types of sport on the questionnaire sent at the age of 17.
The fourth article (IV) studied the association of health-related behaviours, social relationships and health status with persistent exercise and persistent inactivity, as defined using information from all three questionnaires. The information concerning stability of physical activity was based on all three questionnaires; health-related behaviours and health status were based on the questionnaire sent at the age of 16, and school type on the questionnaire sent on the17th birthday. The number of subjects answering all three questionnaires totalled 5028 (2311 boys and 2717 girls), with a response rate among boys of 75.8 % and among girls of 81.7% (III, IV). Of these, 122 twins (57 boys and 65 girls) were excluded because of incomplete answers or due to an illness or handicap that could affect physical activity. The final cohort size was 4906 subjects (2254 boys and 2652 girls) (III, IV).
The very active group: exercise 4-5 times a week or more with profuse or moderate sweating and breathlessness. The active group: exercise 2-3 times a week with profuse or moderate sweating and breathlessness, or exercise 4-5 times a week or more with little sweating and breathlessness. The moderately active group: exercise 4-5 times a week or more with no sweating and breathlessness, or 2-3 times a week with little or no sweating and breathlessness, or once a week with profuse, moderate or little sweating and breathlessness. The hardly active group: exercise 1-2 times a month or less than once a month with profuse, moderate or little sweating and breathlessness, or exercise once a week, 1-2 times a month or less than once a month with no sweating and breathlessness. The inactive group: exercise less than once a month or not at all and no leisure time physical activity.
We also asked own perception of physical fitness with five alternative answers (very good, rather good, satisfactory, rather poor, very poor).
Stability of adolescent physical activity (III, IV)
Other measures of adolescents' sports (III)
Because subjects could be active in more than one sport, we used these three sport groups (aerobic, power and others) to form eight different potential combinations. We then studied the distribution of subjects in these combinations, and the proportions of persistent exercisers and persistently fit subjects therein. Finally, we divided adolescents into those who participated in ball games (football, volleyball, badminton, tennis, baseball, basketball, rinkball, ice-hockey) and those who did not. We also compared those who took part in organised sports with those who did not to assess the social aspects.
Parental and grandparental physical activity (II)
Parents' physical activity
The intensity of physical activity was measured by estimated MET (metabolic rate for physical activity) values. Walking was estimated as 3 METs, rapid walking and a mixture of walking and jogging as 6 METs, light jogging as 10 METs, and running as 13 METs (Ainsworth et al. 1993).
Total MET values were calculated in the first questionnaire by multiplying the time used in physical activity (using the class midpoints) by the estimated MET value and adding the four values together. In the second questionnaire, total MET values were calculated by multiplying the estimated MET value by the duration of one physical activity session and the duration of physical activity per month using the class midpoint, and then dividing by four to obtain weekly values. If the subject answered 'not participating in any leisure time physical activity' the estimated MET value was 0.
Mothers' and fathers' total MET values were classified into quintiles (20% of subjects in each class). Those who answered the first or second versions of the questionnaire were classified separately. The mean age of mothers and fathers was equal with both questionnaire versions. Classifications of physical activity in both questionnaires were then combined. The highest 20% were considered as very active and the lowest 20% as inactive. In some analyses parents in these extreme classes were compared with adolescents' physical activity (II).
Grandparents' physical activity
Health-related behaviours, social relationships and health status
(I, II, IV)
Weight was asked to the closest kilogram and height to the closest centimetre. Body mass index (BMI) was computed as kg·m-2. Use of dietary fats was measured by eliciting the types of spread the subjects use on their bread. In the present study, a detailed dietary history could not be taken.
Breakfast eating habits were also measured with a single question using three alternatives (once a week, 3-4 times a week, daily).
Smoking habits were measured by responses to three questions: on smoking initiation, the amount of cigarettes ever smoked and current smoking habits. Based on these questions four different classes of smokers were formed: (1) non-smokers, (2) occasional smokers, (3) regular smokers, and (4) quitters. Subjects who had smoked one cigarette at most or had never smoked were assigned to the non-smokers class. Those who had smoked a total of 2-50 cigarettes but had never regularly smoked, and subjects who had smoked at least one cigarette or smoked less than once a week were assigned to the occasional smokers class. Subjects who reported having smoked 2-50 cigarettes or more during their lifetime and currently smoked at least once a week were assigned to the regular smokers class. Subjects who reported that they had quit smoking were assigned to the quitters class. Subjects with contradictory answers were excluded from the classification (girls 0.4 %, boys 1.7 %).
Alcohol consumption was measured by three questions: on the frequency of alcohol consumption, the frequency of times intoxicated, and the frequency of slight intoxication. Based on the responses to these questions alcohol use was classified into three categories: (1) non-users, (2) users, and (3) heavy users. Subjects who drank alcohol once a week or more or were drunk at least once a month or slightly intoxicated at least once a week were considered heavy users for their age. Subjects who did not use alcohol and never got even slightly intoxicated belonged to the non-users class. If answers corresponded to other alternatives, subjects were considered to belong to the users class.
Social relationships (I, II, IV)
Parents' socio-economic status (II, IV)
The main socio-economic categories in the second article were self-employed persons, upper-level employees with administrative, managerial, professional and related occupations, lower-level employees with administrative and clerical occupations, manual workers, students, pensioners, and others. In the fourth article the categories were combined into upper-level employees, lower-level employees, workers (manual workers, pensioners, unemployed), self-employed, and farmers. More detailed descriptions appear in articles II and IV.
Health status (I, IV)
(All questions appear in the appendix).
Data analysis and statistical methods (I, II, III, IV)
Log linear models were used to analyse the simultaneous relationship of physical activity to several categorical variables. Thus, in the first article, a log-linear model with terms for all two-way interactions between physical activity, sex and independent variables were fitted first. Then the three-way interaction was fitted and the change in model fit relative to change in degrees of freedom was used as the statistical test (I).
In the second article, to assess familial aggregation of physical activity, intra- and inter-generational correlation coefficients were computed between relative pairs.
In the third article, Spearman correlations were used to assess the strength of associations between ordered variables, while odds ratios and their 95% confidence intervals (OR, 95 % CI) were used correspondingly for pairs of dichotomous variables (III).
In the fourth article, the associations between persistent physical activity and health-related behaviours and other determinants were assessed by logistic regression analyses, which were used to screen which variables showed significant associations with the outcome measures, first individually and then by multivariable analysis within groups (health-related behaviours, social relationships and health status). A final sex-specific logistic regression was done with those variables that remained significant. The possible lack of statistical independence between members of a twin pair was taken into account by using logistic regression modelling with generalised estimating equations providing correct confidence intervals for odds ratios.
Analyses were computed with the SAS program package (version 6.12, Cary,
Distribution of physical activity (I)
Stability and changes of physical activity (I)
When the data was analysed as individuals with repeated measurements, 20.4% of boys and 13.0% of girls were persistent exercisers (had remained in the same highest categories of exercise frequency over the three measurement occasions), and 6.5% of boys and 5.3% of girls were persistently inactive (in the same lowest categories of exercise frequency over the three measurement occasions).
Among boys who considered their fitness to be very good at the age of 16, 43.6% were persistent exercisers and 1.6% persistently inactive, while the corresponding figures for girls were 39.7% and 1.6%. The intensity of physical activity at 16 years also correlated with persistent exercising; of boys who breathed and sweated heavily during exercise, 42.8% were persistent exercisers and 1.3% persistently inactive, while the corresponding figures for girls were 31.5% and 3.9%.
Correlations with other measures of sports (III)
We formed eight groups covering all the possibilities of participating in different groupings of sport types. For this analysis, we also considered all the open-ended answers not included in the 19 given alternatives, and assigned those persons too to the three possible groups (aerobic, power, others). Those participating only in sports in the aerobic group had the lowest proportion of persistent exercisers, while those reporting participation in sports from all three groups were more often persistent exercisers (Table 5). Power exercise also seemed to be associated with persistent physical activity.
An interesting finding was that among boys who participated in ball games, 23.1% were persistent exercisers, while among boys not participating in ball games the figure was 12.6%. Among girls who participated in ball games, 25.0% were persistent exercisers, and among girls not participating in ball games the figure was 8.6%
Overall, 35.2% of boys reported that they participated in organised sport compared to 20.5% of girls. Among persistent exercisers, 80.4% of boys and 61.2% of girls participated in organised sports groups of some kind.
Correlations of health-related behaviours, social relationships and health status with physical activity
Associations between health-related behaviours and physical activity
The use of alcohol differed significantly across physical activity groups at the age of 16 (girls p< 0.001 and boys p = 0.056) (Figure 5). The frequency of heavy users increased systematically as the physical activity level decreased, and the frequency of non-users decreased when the physical activity level increased. Among girls, 9.8% of the very active and 29.7% of the physically inactive were heavy users. Among boys, 14.0% of the very active and 26.1% of the inactive belonged to the heavy user class. When smoking habits were taken into account in a three-factor log linear analysis (smoking, alcohol use and physical activity), alcohol was no longer significantly associated with physical activity. The association between alcohol use and physical activity was thus accounted for by the association of smoking and alcohol use, and of smoking with physical activity. In the three-year follow-up, alcohol use was no longer associated with persistent exercise or persistent inactivity (Table 6).
There were no significant differences in either sex in the body mass index by physical activity level at the age of 16, or by persistent physical activity.
The use of spreads on bread differed significantly across physical activity classes at the age of 16 (girls p = 0.001 and boys p = 0.037). Among girls, the proportion of those who did not use any spread increased when the physical activity level increased; in the very active class 15.4 % did not use any spread compared to 1.6 % in the inactive class. Among boys the corresponding proportions were 5.2 % in the very active class and zero in the inactive class (Figure 6). In the three-year follow-up, too, non-use of spreads on bread was associated with persistent exercising among girls, but not among boys. Also, regular breakfast eating was associated with persistent exercising among boys and girls, but irregular breakfast eating only among persistently inactive boys (Table 6).
Associations between social relationships and physical activity (I,
In a three-factor log linear analysis (physical activity, smoking, school type), both smoking and school type remained significantly associated with physical activity among girls and boys at the age of 16. No significant difference was found in the leisure time company in different physical activity groups, nor between boys and girls. In general, boys (41 %) seemed to spend more time in large groups than girls (31%).
In the three-year follow-up, mother's socioeconomic status was associated with persistently active boys, and father's with persistently inactive boys, but there were no associations among girls. The final logistic regression results are shown in Table 6.
Associations between health status and physical activity (I,IV)
Eleven percent of the girls and 10 % of the boys reported a long-term
illness that hindered daily activities. The five most common reasons were
allergies, asthma, errors of refraction, diabetes mellitus and migraine.
Familial aggregation of physical activity (II)
Among all girl pairs, 69.6% of very active girls had a very active sister
and 45.5% of inactive girls had an inactive sister. Among all boys, 58.7%
of very active boys had a very active brother and 39.4% of inactive boys
had an inactive brother.
Grandparents' own physical activity patterns, as reported by their adult children, showed that 55.5% of physically active maternal grandmothers had an active husband and 17.9% an inactive husband. In contrast, only 10.8% of inactive maternal grandmothers had an active husband and 64.2% of inactive grandmother had an inactive husband (correlation 0.43). Among active paternal grandmothers 43.7% had an active husband and 25.3% an inactive husband. Among inactive paternal grandmothers 11.5% had an active husband and 66.3% an inactive husband (correlation 0.33).
Intergenerational physical activity patterns seemed not to be associated, as these correlations were quite low in the entire family data. The correlations between father and adolescent (0.08) and mother and adolescent (0.09) were similar. The strongest correlation was between father and boy (0.10), and weakest between father and girl (0.05). Grandparental physical activity was not associated with the parental or adolescent physical activity patterns. Because of the lack of associations across compared to within generations, we examined how persons at the extremes of the distributions of physical activity were associated. Table 7 shows cross-tabulation of the extreme classes (very active and inactive) of parents and children.
There was a significant difference between very active and inactive mothers
and their daughters' physical activity (p<0.001), but not between extreme
classes of mothers and their sons (p=0.142). No statistically significant
association was found between extreme classes of fathers and their daughters
(p=0.898) or sons (p=0.096).
Stability and changes of physical activity
Some longitudinal studies have found moderate to low tracking correlations, and others moderate to high. In a Finnish six-year follow-up study (baseline ages 12, 15 and 18 years) all tracking correlations over a six-year period of physical activity were significant, varying from 0.5 to 0.8 among boys and 0.4 to 0.61 among girls (Raitakari et al. 1994). Tracking correlations are generally moderate or low (Kelder et al. 1994, Anderssen et al. 1996, Pate et al. 1996, Telama et al. 1996, 1997), but according to Janz et al. (2000) tracking of physical activity was moderate to high. In the Finnish study mentioned above, physical inactivity showed better tracking than physical activity, and subjects who were constantly inactive had a less beneficial coronary risk profile (Raitakari et al. 1994). Similar results were found in the Muscatine study, where sedentary behaviour tracked better in boys and vigorous behaviour in girls (Janz et al. 2000).
Those who are persistent exercisers during adolescence may be more likely to continue a physically active life in adulthood, where the health benefits will be seen. This research project shows that healthier lifestyles are associated with higher physical activity, and therefore the persistently inactive group is the biggest challenge in promoting health. However, this does not mean that inactive adolescents are doomed to be inactive adults.
Correlations with other measures of sport
An association of organised sports with high physical activity stability has been reported earlier. Participation in competition sports (Telama et al. 1997) and membership of a sports club (Barnekow-Bergkvist et al. 1998) are predictors of later physical activity. According to Telama et al. (1996), the tracking of participation in organised sport is moderately high among adolescents compared with the dropout rate from sports in general. This may be because the stability of non-organised physical activity appears to be lower than the stability of organised sports. The explanation for this could be that it is more difficult to estimate and remember non-organised activities (Telama et al. 1996). Also, the role of organised sports might change during adolescence. According to a 15-year follow-up, over the course of time organised sports activities became a relatively more important contributor to weekly habitual physical activity (HPA) and energy expenditure (van Mechelen et al. 2000).
Based on this study, among boys the proportion of persistent exercisers was highest in those who participated in cross-country skiing, jogging and bodybuilding, and among girls in those who participated in ball games. Whether these historical trends hold as new sports increase in popularity and spread among adolescents remains to be seen; recommending one particular sport over another should be done very cautiously. The association of power type sports and ball games with persistent physical activity is also interesting. Is power type exercise a supplementary training to some other sports, or are there some other reasons, for example friends at the gym? Power types of exercise can be forms of both personal and social training. Is participation in ball games equivalent to organised sports? These issues need further investigation.
Association with health-related behaviours, social relationships and
In the Finnish cohort study (ALHS), students involved in interscholastic sports were less likely to be regular or heavy smokers than others who had not participated (Rimpelä and Rimpelä 1993). A negative correlation between physical activity and smoking has usually been reported to be stronger among boys than girls, but our study showed no gender difference. Among adolescent girls, smoking is considered a good way to lose weight. Dieting among girls may also exacerbate the risk of initiating smoking (Austin and Gortmaker 2001). Therefore physical activity should be recommended more as a healthy way to control weight.
Alcohol use and abuse
The results of the very few studies of physical activity level and alcohol use among adolescents differ. According to a Finnish study of 11-16-year-olds, those who used alcohol spent less time in physical activities than non-users. Girls in the drinking group were less active than boys (Honkala 1984). Page et al. (1998) studied a nationally representative sample of 12 272 US high school students and found no association between alcohol consumption and participation in team sports. An association of alcohol use with reduced participation in conventional leisure-time activities in early adolescence was found in another US study of 1 092 predominantly low socio-economic status rural school children aged 11-18 (Terre et al. 1990). Opposite results were reported among 878 college students: 80% exercised regularly and consumed alcohol (Blank et al. 1993).
The association of physical activity with alcohol use is not clear; it might be associated with developmental, pubertal, adolescent or environmental factors. Our study found the alcohol association was related to smoking. According to Felton et al. (1999), alcohol consumption was associated with parental or friends' drinking habits and increased with age. Regional aspects might also affect drinking habits due to socio-economic factors (Karvonen and Rimpelä 1998). Physical activity, especially in team sports, might simultaneously promote social bonding and increase the sense of belonging, thereby lowering risk-taking behaviour. Sports events are often sponsored by alcohol or beer companies, so it would also be interesting to study whether alcohol sponsorship has any effect on adolescent behaviour.
In this study, mean body mass index was not associated with physical activity among boys or girls. Less than 10% of the boys and girls had a body mass index (BMI) over 22, so there appears to be few overweight adolescent 16-year-olds. Kemper et al. (1999) found that high physical activity in both sexes was related to low body fat. Also, according to Klesges et al. (1984), obese children are less active than those of normal weight, but a Finnish study of school children found no significant association between low physical activity and high BMI (Telama et al. 1983). Van Mechelen et al. (1999) found a positive longitudinal relationship between HDL-cholesterol and physical activity, but low tracking was found for physical activity and various dietary intake variables.
An association of low socio-economic status (SES) with low physical activity has been reported elsewhere. Low socio-economic status was associated with less physical activity in other studies (Lee and Cubbin 2002), but Gordon-Larsen et al. (2000) found in a representative US sample that maternal education was inversely associated with high inactivity of high school adolescents. Also, according to Kristjansdottir and Vilhjamsson (2001), upper class students were less sedentary and more physically active than working class students. In a Danish study, 370 children aged 6-18 were followed-up for 13 years: lower physical activity was associated with low education among young women, while father's activity level at work was associated with young men's physical activity. Men whose parents reported high physical activity at work were less often physically inactive during leisure time (Osler et al. 2001). According to Daley and Ryan (2000) no significant association was found between physical activity and academic performance among 13-16- year-old English adolescents. A Finnish study found that low parental socio-economic status was associated with smoking, low physical activity and obesity among boys (Leino et al. 1996).
Our study project found associations of school type and school achievement with physical activity. Comprehensive schooling ends at the age of 16, and adolescents then have to decide about their further education. At the same time adolescents are making many other decisions about their leisure-time hobbies, with whom they spend their leisure time, how they spend their pocket money, etc. Late adolescence can also be the time of life when decisions are made about physical activity. In this project we did not study the social aspects of late adolescent decision-making. Social relationships encompass a broad area, of which this study covers only part.
Familial aggregation of physical activity
When the overall data set was analysed, the 16-year-old adolescents' physical activity patterns were not associated with either parents' or grandparents' physical activity patterns. When the extremes of physical activity patterns were considered, a significant association between mothers and daughters was seen. The corresponding association for fathers and sons was not as strong, but the results suggest that parents are more likely to affect the physical activity patterns of their same-sex children in adolescence, while opposite-sex offspring are less influenced. Also, very active or quite inactive parents appear to affect their children more than moderately active parents. However, it cannot be totally excluded that in families with adolescents aged 16 years, adolescent physical activity patterns influence parental habits. For example, a physically very active youth may activate a previously less active parent into more physical activity.
Among other studies, Stucky-Ropp and DiLorenzo (1993) found certain variables to predict the level of physical activity among boys and girls in the 5th and 6th grades. Support and modelling from friends and family were predictors among boys, but not girls. For girls only, physical activity appears to be more highly influenced by the number of exercise-related pieces of equipment at home, and parental modelling of exercise. Father's occupation is significantly related to the frequency of exercise (Gottlieb and Chen 1985).
Although our study population consisted of twins, we generally discarded their twinship and did not study genetic associations. Genetic association was reported by Beunen and Thomis (1999), who found that if one of the parents or co-twins is active in sports, it is more likely that the child or co-twin is also similarly active. According to another twin study, monozygotic twinpair correlations of physical activity are reportedly higher than dizygotic, suggesting that genes play a role in physical activity patterns (Lauderdale et al. 1997).
This study did not include genetic aspects nor take causality into consideration. Longitudinal studies of children through adolescence and of their parents would be necessary to examine the pattern of causation for these observed associations.
Furthermore, the gender differences are pronounced in most countries. For example, the proportion of girls exercising actively is approximately half that of boys among 15-year-olds in Greenland, Lithuania and Greece (Hickman et al. 2000). Also, according to Portuguese study, girls belonged more often to inactive or low active groups than boys (Mota and Esculcas 2002). The issue of gender differences is also relevant when we consider physical activity tracking from adolescence into adulthood. According to Barnekow-Bergkvist et al. (1998) more men than women participated in sports activities (70% vs.41%) at the age of 16, but at 34 years there was no significant difference.
Health-related behaviours and other determinants
Girls seem to be a key target group for the prevention of both smoking and sedentary life styles. Gender differences are pronounced in most studies, but for unclear reasons. One reason might be that boys are more likely to participate in ballgames and organised sports. The situation might be different today, since new types of physical activity, e.g. aerobics, have become very popular among girls, at least in Finland. Puberty, nutritional habits and environmental factors such as improved facilities might all have influence, and should be studied further.
Our study had a high response rate and we have earlier shown a high one-month test-retest reliability of our questionnaire items. The questionnaire-reported physical activity correlated moderately with the laboratory tests, interview and VO2max measurement. There are contrary opinions about the use of questionnaires for such a purpose. According to Sirard and Pate (2001), subjective techniques such as surveys, self-report questionnaires, interviews, proxy-reports and diaries are the least reliable methods. However, subjective methods are often the only cost-efficient way to study large samples in epidemiological studies.
Previous studies have also shown that a single question concerning leisure-time physical activity correlates with maximal oxygen uptake and can be used to provide useful information about fitness and physical activity (Schechtman et al. 1991, Siconolfi et al. 1985). Crocker et al. (1997) Considered that a physical activity questionnaire for older children is a cost-efficient method of assessing general levels of children's physical activity during their school years. Valid and appropriate measurement of physical activity is, however, a challenging task because it can vary considerably both within and among individuals and populations (Kriska and Caspersen 1997), and because there are several health-related dimensions of physical activity, such as caloric expenditure, aerobic intensity, weight bearing, flexibility and strength (Caspersen 1989).
In this study concentrating on adolescent physical activity patterns we decided to use a self-report questionnaire as the measurement tool; the number of subjects was so large that this method was the most cost-efficient way to study physical activity. The same method has been used in another Finnish study, the biannual Adolescent Health and Lifestyle Survey (AHLS) of adolescent health and lifestyle among 12-, 14-, 16- and 18-year-olds. In the AHLS two variables are used to measure physical activity: frequency - by asking the amount of time spent in physical activity, and intensity - by the level of sweating and breathlessness. In our study we formed physical activity categories by combining these two questions.
We repeated the same questions three times over a three-year period to examine changes in physical activity patterns. However, the response rate among those answering the first questionnaire was higher than among those who answered all three. We did not study whether those answering all three were more active in general, and the drop-out effect could not be estimated in this work.
Although our study population consisted of twins, we discarded their twinship in the analyses and considered them solely as individual adolescents drawn from the population. This procedure has been used in earlier analyses of individuals of the twin cohort (Kaprio et al 1990, Verkasalo et al. 1997), and inferences from epidemiological analyses of twins considered as individuals are comparable with those based on singletons. We have shown that physical activity patterns in twin individuals and singletons are similar (II). There is little reason to assume that the determinants of change in physical activity patterns would be different in large numbers of twins or singletons.
The material of this study is unique, as physical activity patterns of three generations have not been reported in earlier studies. Information on grandparental leisure time physical activity was based on parents' reports, which might raise the question of its reliability. The lack of association of three generations of physical activity might be due to the measures we used, which differed between the generations. The parental and adolescent measures of physical activity reflected current activity patterns, which may be highly age- or cohort-specific. On the other hand, the assessment of grandparents was based more on a global lifetime perspective. Hence, these may not be fully comparable if physical activity patterns vary a great deal over time. We therefore based our analyses on computing relative activity levels within generations.
Sociocultural changes in Finland
Due to the growing awareness of the health benefits of physical activity, leisure-time physical activity has gained higher priority as a means of keeping fit and preventing the possible negative consequences of physical inactivity.
Moreover, the opportunities for leisure-time physical activity have expanded from generation to generation. Today's adolescents have access to numerous alternative forms of exercise. At the end of the 1950s there was one physical activity facility for every 1200 citizens, of which two-thirds were outdoors, whereas in 1990 there was a sports facility for every 200 citizens (Heikkinen et al. 1992).
As these options have expanded, physical education classes at school have decreased. Today, the responsibility for physical activity falls more on the adolescent him/herself, or their family. Nevertheless, comprehensive schools need to recognise that there appear to be individual differences in the development of high-risk behaviours, and that scholastic aptitude seems to be an indicator of risk. Promoting physical activity among adolescents should lead to healthier overall lifestyles.
Physical activity recommendations
This work did not examine whether adolescents participate in sufficient physical activity. In the study of Nupponen and Telama (1998), one third of the boys and one fifth of the girls had a sufficient physical activity habit when the criterion was at least four times a week, a total of four hours. Further studies of physical activity are needed in order to make scientifically based recommendations for Finnish adolescents.
Knowledge of the health benefits of physical activity and the risk of declining activity in adolescence makes young adolescents an important target group for physical activity promotion programmes.
Based on this study, physical activity patterns during adolescence show changes in behaviour over a three-year period. Those participating in organised sports, ballgames and power types of sports appear to be more persistently active.
Health-related behaviours, especially smoking, as well as the type of school and educational achievement, have an impact on persistent activity. Physically active adolescents seem to live a healthier life and tend to be better educated.
Familial influence is mainly based on siblings' or co-twin's physical activity. Boys seem to be more active than girls.
Based on our findings, schools emerge as the key elements in promoting physical activity. The role of the parents decreases in late adolescence and friends become more important. Schools have the possibility to apply systematic physical education and therefore adolescents approaching the end of secondary schooling can be considered a potential "drop out group".
Girls should be a special target group for promoting physical activity during late adolescence, including those who do not take part in organised sports. Another very important target group for promoting health are persistently inactive youngsters. We should focus on finding new and appealing ways of organising physical activity for children and adolescents who are relatively inactive and perhaps have less talent for competitive sports. My opinion is that schools should not only increase physical education, but also improve the quality of classes on offer, in order to make them more attractive. This is one way to encourage participation in more organised sports.
Promoting physical activity among adolescents should lead to a healthier overall lifestyle. Any short-term economic gains achieved by cutting back on physical education classes will be more than lost forty years on in the form of increased health care costs.
This study was carried out at the Department of Public Health (KTTL), University of Helsinki, with the co-operation of the Department of Physiology, University of Kuopio. I wish to thank KTTL for the privilege of using the extensive Finn Twin datasets gathered there. This study received financial support from the Ministry of Education for the first years. This support is gratefully acknowledged.
I have been very fortunate to have two excellent supervisors, professor Jaakko Kaprio and docent Urho Kujala. They always found time to read and comment on my manuscript during these years, which in my case involved a lot of time! Their constructive comments, continual support and endless patience have been vital to the success of this study. I also thank docent Heikki Pekkarinen for his encouragement to start this work and his support throughout these years.
The two official referees of the dissertation, docent Olli Raitakari from the University of Turku and docent Sakari Karvonen from Stakes, are gratefully acknowledged for their careful work and constructive criticism that helped to clarify the manuscript during the final stages.
I wish to express my gratitude for all the statistical and technical assistance I have received with this work. Special thanks to Torsten Winter for his time and patience with the statistical analysis, to Lauri Karppinen for helping with the pictures, and to Richard Burton for the language revision. I also thank my Wellmedia colleagues for their support, and for understanding the time this work has taken from my daily work. I particularly thank my business partner Arto Tuominen for his way of motivating me by challenging and endless encouragement during these years.
I am sincerely grateful to all my friends for their encouragement and good company. I especially thank them all for reminding me of the importance of other interests in life.
Warmest thanks go to my family. My parents, Pirkko and Pentti Knuuttila, have shown such confidence in me, along with endless support and interest in my work and well-being. Their appreciation of my academic studies has encouraged me throughout these years. I am also deeply grateful to Kari for his support, and for sharing with me these most challenging years, and to our wonderful children, Sebastian, Hans-Christian and Charlotta, for being the sunshine of my life.
Helsinki, November 2002