6.6. Methodological considerations

The major strengths of this prospective study are the large and unselected population sample, the opportunity to control for potential confounding factors, a long follow-up time and a high participation rate. Extensive data based on the general population offered a unique opportunity to study the association between a wide range of adolescent sports and adult physical activity. The measurement at 31 years can be used as baseline measurement in later follow-ups, in which the level of physical activity and fitness at 31 years will be related to incidence of diseases in later life.

Reliance on self-reported measures of physical activity is a limitation in surveys of this kind. Social desirability bias can lead to over-reporting of physical activity (Sallis & Saelens 2000), and therefore the number of inactive individuals may actually be greater than that reported. Such an obviously non-differential bias is, however, unlikely to affect the association between physical inactivity and the explanatory factors. Our questions about physical activity status in adolescence should be interpreted as indicators of engagement and participation in physical activity. We did not measure the total amount of physical activity in adolescence, but only participation in sports, which is not always the predominant amount of physical activity during adolescent years. Questionnaire methods do not generally provide accurate data on individual energy expenditure, but they are considered useful for grouping people into categories on the basis of their physical activity (Montoye et al. 1996).

Selective losses to follow-ups may be an important source of bias in this kind of longitudinal studies. The compliance with the study may be higher in the healthiest part of the population. In the present study the response rate at 14 years was extremely high and enabled the characterization of those who did not participate in the follow-up at 31 years. Those who did not participate at all in the follow-up at 31 years did not essentially differ from those who participated at 31 years with regard to key variables, such as physical activity and body mass index at age 14, although they came slightly more often from families with a low social class and had slightly more often low grade average at school at 14 years. Those who participated in the medical examination at 31 years were not physically more active or less obese compared to those who only answered the postal inquiry at 31 years. The population that participated in the follow-up at age 31 seems to be representative of the whole Northern Finland birth cohort of 1966 by their level of physical activity and BMI.

The highest level of evidence is generally thought to be achieved from randomized controlled trials. However, some phenomena cannot be evaluated in trials of this kind. For instance, it would not be ethically acceptable to form a study group of children who would remain physically inactive until adulthood and to compare them with a physically active group. The present study includes cross-sectional and prospective observational research. The causality between the variables cannot be fully asserted based on these kinds of study settings. Therefore, the results should be interpreted with a certain amount of caution, also keeping the possibility of resiprocal effect and natural selection in mind.

The results can most likely be generalized to represent Finnish adults aged about 31 years, as well as other Western developed countries with similar geographical and sociopolitical situations. Some results, for instance the prevalence of different types of sports are typical for the area of northern Finland. However, the society and also the types of sports among adolescents have changed considerably since 1980, and further information is required to evaluate the patterns and the social determinants of adolescent physical activity today.