The evolution of physical activity as an exposure

Physical activity as an exposure has made the same process of evolution as many other exposures in the field of epidemiology. When little is known, simple, unsophisticated and inexpensive methods are used to demonstrate disease associations [1]. Physical activity was initially measured with rough methods and showed early on an association with disease. One of the first epidemiological studies on the impact of physical activity was conducted in the 1950s. The authors found that drivers of London’s double-decker buses were more likely to die from “coronary thrombosis” than the more physically active conductors [2]. Assessment of physical activity during working hours kept its position as a primary way of measuring physical activity for many years; although, studies solely based on work-related physical activity have numerous drawbacks. Today a full-time worker spends only roughly 20% of the total hours in a year at the workplace. As a result of modernization, physical activity levels at work continue to decline, leaving other types of physical activity to be a better representation of the total picture.

The issue of measurement changes over time. During the last decades research has moved from description of simple relationships to more advanced questions such as: which aspects of physical activity are of importance? What level of physical activity is necessary to stay healthy? Does timing matter; does physical activity in different life stages have different impact on disease? Progressively, research has focused on related issues as well, such as: which factors affect the habits of health-enhancing physical activity? These questions result in a need for more accurate exposure quantification.

Available instruments from the era of the descriptive phase do not always meet the demands of the more advanced phase. New improved, validated and reliable methods are needed and continue to appear.

Given the importance of the field, it is very satisfactory to see the attention the EJE has given to the research field during the recent years. The articles published in the EJE during the last 2 years can be divided into three groups:

  • methodological articles

  • articles with physical activity as the main exposure and

  • articles with physical activity as a covariate.

Methodological articles

The journal has published a number of articles on epidemiological measurement issues. Articles about the evaluation of questionnaires in the area of physical activity have particularly received attention. Several questionnaires have been validated against objective methods such as accelerometers, position, and motion instruments. Contrary to many single-item questionnaires, lacking the ability to capture all activities during everyday life, these questionnaires have generally aimed for assessing total physical activity during all 24 h. Physical activity has been expressed as MET-hours/day (METs representing multiples of basal metabolic rate) or PAL (a weighted average of daily MET hours), units facilitating comparability between studies.

For example, Orsini et al. studied the validity of a total physical activity questionnaire among 116 elderly women using both accelerometers and a 7 day physical activity diary. Deattenuated concordance correlations were 0.38 respective 0.64 [3]. Reliability was studied with a 1 year test-retest in the same cohort. The intraclass correlation coefficient for current physical activity was 0.69 [4]. Matthiessen et al. validated a Danish adapted version of the IPAQ (International Physical Activity Questionnaire) with a body position and motion instrument calculating energy expenditure, among 138 men and women aged 20–59. Spearman’s rank order correlation between methods was 0.71 [5]. Schmidt et al. evaluated a short questionnaire for usual physical activity in the distant past. Two months after a larger “reference interview” 110 breast cancer cases and 101 controls between 50 and 74 years of age were reinterviewed. The mean difference between the two questionnaires was 3 MET hours/week [6]. In addition, Graff-Iversen et al. conducted an indirect validation study of two short leisure-time physical activity questionnaires with serum lipids, anthropometric measurements, and aerobic power among 1,238 men and 1,550 women. The Spearman’s rank order correlations were between 0.1 and 0.3 [7].

Earlier epidemiological studies of physical activity often assumed that the instrument used automatically was valid and reliable. These studies may have run the risk of presenting spurious results due to poor quality measures. Today, however, evidence of validity and reliability is becoming standard. Although a method has been validated, some attention should be given to how the validation study was designed, and in which population the method was validated. For example, the sample chosen in the validation study may not be representative for the group where the instrument is planned to be administered.

Articles with physical activity as the main exposure

Moving to the phase of more advanced questions, EJE has published a number of studies with more sophisticated questions pertaining to different facets of physical activity. As physical activity takes many forms, it can also be measured in a variety of ways. The choice of method in any given study depends on a range of factors. There is a clear diversity of methods used in the following examples of articles published in the journal during 2006–2008.

Which aspects of physical activity are of importance?

Huth Smith et al. used a self-administered questionnaire with two questions about specific types of physical activity. They conducted a cross-sectional study in a population-based random sample of 6,784 Danes. Commuting physical activity, independent of leisure time physical activity, was associated with a healthier level of HDL and LDL cholesterol, triglycerides, BMI, and waist circumference [8]. Tulmunen et al. measured maximal oxygen uptake (VO2max) during a cycle ergometer test in 1,519 middle-aged Finnish men. Poor cardiovascular fitness was associated with having elevated depressive symptoms [9].

What level of physical activity is necessary to stay healthy?

Aadahl et al. found a significant linear association between overall self-reported physical activity up to 45 METs/day and BMI, waist circumference, waist circumference, HDL and triglycerides in 1,693 men and women 33–64 years of age. Limited gain, in terms of cardiovascular risk factors, seem to be accomplished by an increase to a higher level of physical activity [10]. In the commuting study above, spending more than an hour a day on commuting physical activity was favourably associated with biological risk factors for cardiovascular disease [8].

Does timing matter; does total physical activity at different ages in life have different impact on disease?

Wiklund et al. tried to answer this in a population-based case-control study comprising 1,449 prostate cancer cases and 1,118 controls. The self-administered, written questionnaire Orsini et al. validated above [3] was used. The study did not reveal any statistically significant associations. Similarly, there were no significant associations between changes in activity over time (increased, decreased, or continuously high activity as compared to continuously low activity) and risk of prostate cancer [11].

Which factors affect the habits of health-enhancing physical activity?

In the area of genetic epidemiology, family clustering in physical activity was studied among 2,375 Portuguese families. The self-administered Baecke questionnaire was used [12]. This questionnaire has five-point Likert scales, ranging from never to always or very often, to map different sections of physical activity. The authors concluded that shared environmental factors appear to contribute more to physical activity habits than shared genetic factors [13].

Articles with physical activity as a covariate

With the accumulating evidence of the association between physical activity and many chronic diseases, physical activity has not only become a main exposure variable in many studies, but also an important covariate. This raises the need for valid comprehensive assessments of physical activity, instruments that are time-effective for increased feasibility.

The EJE has lately published a number of articles with physical activity as a covariate. They have covered diverse topics: the association between father’s social class and adult obesity [14], periodontal disease and risk of myocardial infarction [15], and behavioural risk factors in different generations of non-Western migrants [16]. These studies are some examples where physical activity questionnaires have been used.

Old methods and new opportunities

Questionnaires are easily administered and distributed. They are convenient to the study subject and do not alter the behaviour of the subject being surveyed. They can provide information on a number of lifestyle factors for a large number of subjects. The investment in time and money is lower compared to other methods. Consequently, questionnaires have been the method of choice in larger epidemiological studies. But there is a wide variety of questionnaires and questions investigating different aspects of physical activity to choose from. There has been a lack of sophistication in the measurement of physical activity in many epidemiological studies. Misclassification, lack of precision, and social desirability bias may all lead to inconsistent results between studies. Exposure-measurement error is often ignored when interpreting epidemiological study results [17]. Especially in the area of physical activity research, we would all benefit if it was not overlooked, but rather discussed and if possible assessed.

Light weight data loggers measuring physical activity as steps or as movement intensity overcome some of the inherent limitations of self-reported physical activity. They are free from random and systematic errors introduced by respondents and interviewers, cultural tradition, and language. With the rapid advancements in technology, devices such as accelerometers have become more and more suitable for epidemiological research. However, the raw data provided by accelerometers are in the form of activity counts (ct) reflecting the frequency and the amplitude of activity in a given interval (epoch per time unit) [18]. To translate this into meaningful information, a calibration process is needed. The translation from activity counts into meaningful information is not a straightforward process, and there are considerable differences in the outcome produced from different types of accelerometers. Furthermore, there is no consensus on the cut points used to define physical activity intensities. Hence, comparability is reduced [19].

Several large-scale studies assessing physical activity with objective methods such as accelerometers have been conducted during the recent years. In the European Youth Heart Study with about 1,000 school children per participating centre, physical activity was assessed with accelerometers. High levels of physical activity were significantly associated with lower body fat [20] and cardiovascular risk factors [21]. In the Swedish ABC-study 2002–2003 and the US National Health and Nutrition Examination Survey (NHANES) 2003–2004, accelerometer data were obtained from thousands of individuals representative of the population [22, 23]. Interestingly, compliance with current physical activity recommendation according to accelerometer-measured activity was considerably lower than according to self-reported activity.

In epidemiology, objective methods have mainly been used to validate other activity assessment methods such as questionnaires—a subjective method. Nowadays, data loggers are becoming more affordable and increasingly reliable. When logistical obstacles of delivery and retrieval can be beat, epidemiology may have another method to choose from for the assessment of current levels and patterns of physical activity. Nonetheless, as epidemiologists we may still want to assess past physical activity—which real time measurement will not give us. Thus, questionnaires will still be a part of the future physical activity assessment repertoire.

Physical activity has been measured in a number of ways in epidemiological, interventional, and experimental research. Assessment has remained the weakest link in rigorous attempts to quantify the effects. With increased demands, continuous development and cross-fertilization—between various disciplines and perspectives—the evolution of physical activity as an exposure can continue and lead to new insights. As Kelvin remarked, to measure is to know.