Click on the links below to access all the ArticlePlus for this article.
Please note that ArticlePlus files may launch a viewer application outside of your web browser.
Physical activity has been associated with a reduced risk of several chronic diseases, including diabetes,1 coronary heart disease,2 and colon cancer.3–5 Physical activity could decrease the risk of disease by leading to weight loss or maintenance of a healthy weight, as well as through effects on immune function, metabolic hormones (such as insulin and glucose), or growth factors.4 It is likely that physical activity exerts its strongest effect when maintained over many years, suggesting that long-term activity is more important than recent activity, particularly for diseases such as colon and breast cancers that take years or decades to develop. Although there is considerable published research on methods to assess current or recent activity,6,7 assessment of long-term activity has been less well studied and poses additional challenges.
Measurement of physical activity is complex because the researcher needs to assess the types of activities an individual performs and the frequency and duration that each activity is performed. Furthermore, the intensity of an activity can vary both by type of activity (eg, walking versus running) and by effort of the participant (eg, pace of walking or running). Over time, a person’s activities can change in their duration, frequency, and intensity. In epidemiologic studies, one must balance collection of information on these details with the need to limit participant burden. The questionnaire needs to be simple (eg, for self-administration) and/or short, because physical activity is often not the main exposure of interest, but rather an important covariate.
We developed a one-page recreational physical activity questionnaire to capture data on usual recreational physical activity during the preceding 10 years. This form collected sufficient detail on type, frequency, and duration of activity to estimate the average metabolic equivalent task (MET)-hours (an estimate of the kcals of energy expended per kg of body weight) per week. Specifically, a MET is the ratio of the metabolic rate for the specific activity divided by the resting metabolic rate (which for an average adult is approximately 1.0 kcal/kg per hour).8 For example, a 3-MET activity (eg, walking 2 to 3 miles per hour) requires 3 times the metabolic energy expenditure of sitting quietly. We regarded a 10-year history as long enough to influence chronic disease risk but short enough to be practical for the respondent. This questionnaire is part of the Vitamins and Lifestyle (VITAL) study, a cohort study to investigate the association between dietary supplements and cancer.
Although there is a wide range of validation measures (eg, diaries or physical activity monitors) for current or recent recreational activity, there are no feasible validation measures for long-term activity. Therefore, we administered a more detailed in-person interview covering activity over the preceding 10 years as a comparison measure. We also used current body mass index (BMI; weight in kilograms divided by height in square meters) as an indirect physiologic measure of long-term recreational physical activity, because BMI should be inversely related to average energy expenditure.
The VITamins And Lifestyle (VITAL) Study
Participants in this measurement study were members of the VITAL study cohort. Recruitment of the VITAL cohort began in October 2000 and ended in October 2002. Through April 2002, 61,587 men and women age 50 to 75 were recruited by mail within the 13 counties of western Washington state using a cover letter that targeted supplement users. Respondents completed a 24-page baseline questionnaire that inquired about vitamin, mineral, and herbal supplement use over the preceding 10 years, as well as information on other cancer risk factors, including dietary intake, physical activity, demographic characteristics, and self-reported height and weight.
Measurement Study Participants, Recruitment, and Protocol
The measurement study sample consisted of randomly selected VITAL participants who completed the baseline questionnaire between October 2000 and February 2001 and who lived in the Seattle metropolitan area (King County, WA). We made this restriction so that the study procedures could be administered at the participant’s home. Approximately 3 months after completion of the baseline questionnaire (January through June 2001), participants were asked to take part in the measurement study. Because one aim of the measurement study was to evaluate the relation between dose of supplements and biomarkers, we randomly oversampled high users of vitamin C, vitamin E, and calcium. The results of this component of the measurement study are reported separately.9 We excluded potential participants who reported having Alzheimer’s disease, insulin-dependent diabetes, or any other conditions that would prevent collection of a fasting blood sample.
Participants were asked to complete an in-person interview that included questions about their recreational physical activity and supplement use. The study was approved by the Institutional Review Board of the Fred Hutchinson Cancer Research Center.
Of the 290 eligible participants contacted, 217 (75%) completed the study. Reasons for nonparticipation included inability to be contacted by telephone (9%), no interest (13%), and inability to collect a blood sample (2%). In addition, 1% of eligible participants were excluded because they did not complete the physical activity questions from the baseline questionnaire. The final sample consisted of 112 men and 105 women. Thirty percent of this final sample had been selected because of their high, long-term use of vitamin C or E, or calcium (specifically, those who consumed ≥1000 mg for ≥5 years of vitamin C or calcium, or ≥800 IU for ≥5 years of vitamin E).
One-Page Questionnaire of Physical Activity
The one-page physical activity questionnaire was part of the VITAL 24-page self-administered mailed questionnaire. There were questions about 3 specific recreational physical activities (walking, lifting weights, and yoga) and 2 broader categories of activities (mild and moderate or strenuous exercise) over the preceding 10 years (see the appendix, available with the electronic version of this article at www.epidem.com). Respondents were instructed to report only activities done regularly (at least once per week for at least 1 year during the previous 10 years), and not to include gardening, housework, or activity on the job. Because the questionnaire form was designed for optical scanning, all responses were categorical. To determine the types of moderate or strenuous activities performed, participants were presented with a list of 10 activities from which to select the 1 or 2 activities they did most often. For each activity, participants first reported the number of years in the last 10 during which they undertook each activity, followed by the frequency (days per week) and duration (minutes per day) (see the appendix for categories). For walking only, participants reported their usual pace.
For each activity, the corresponding intensity code (MET) was assigned using a compendium of physical activities that is commonly used to quantify energy costs of various activities in epidemiologic studies.8,10 Similar to Jacobs and colleagues, we calculated energy expenditure (MET-hours) independent of body size.11 Usual MET-hours per week for each activity averaged over the preceding 10 years was calculated as follows:
[Frequency of activity per week × minutes per session × years in the past 10 × MET for activity] ÷ [(60 minutes/h) × 10 years]
We then summed the MET-hours per week for all activities to calculate total MET-hours per week.
Activities were also categorized by intensity level. Specifically, we calculated MET-hours per week of high-intensity activities (ie, activities with a MET of 6 or more), moderate- and high-intensity activities combined (activities with a MET of 4 or more), moderate-intensity activities (from 4 to 6 METs), and low-intensity activities (MET less than 4).12 Moderate- and high-intensity activities were included as a combined group because these types of activities might be more strongly associated with certain diseases than lower-intensity activities.
To calculate these exposure variables, we made several assumptions. Frequency, minutes, and years were calculated as the midpoint of the range of values for a given category (eg, 1.5 days for 1 to 2 days per week). Minutes per session were estimated to be 180 and 65, respectively, for the highest categories of duration (“3+ hours” for low-intensity activities; and “60+ minutes” for all other activities). If only 1 or 2 of the 3 subquestions (years, days per week, or minutes per episode) were answered, we imputed the modal value of the same age/sex strata (for men and women 50 to 64 years and 65 to 75 years) from the entire cohort. Among those who reported doing an activity in the cohort, 2% had missing years, 8% had missing frequency, and 11% had missing duration.
The detailed physical activity comparison interview was modeled after the Minnesota Leisure Time Activity questionnaire.13 We modified the Minnesota questionnaire by focusing on recreational activities, asking about the number of years an activity was performed, and collecting information on pace or intensity.13 Interviewers provided a list of 36 activities grouped into 8 general categories (walking and running/jogging, conditioning exercises, water activities, golf, sports, winter activities, dancing, and other recreational physical activities). Participants identified activities that they had done at least 12 times per year for at least 1 year in the past 10 years. For each activity selected, interviewers asked the age when the participant had started and stopped, and any intervals during that time when they stopped the activity for at least 1 year. Using an open-ended format, participants estimated the usual number of months per year, times per week or month, and minutes per episode for each activity. For walking and running/jogging, participants estimated about how fast they walked (same categories as one-page questionnaire; see appendix) or ran/jogged (casual [>12 minutes per mile], moderate [9 to 12 minutes per mile], or fast [<9 minutes per mile]). For other activities that vary substantially in energy expenditure by effort (eg, bicycling), participants reported their intensity as casual, moderate, or vigorous. If participants made a substantial change in the frequency, duration, or intensity of an activity during the 10-year period, separate entries were made to describe their usual activity pattern during each time period. Interviewers entered the data directly into a laptop computer at the participant’s home. We assigned the MET value for each activity, based both on the type of activity and effort,8,10 and computed MET-hours per week similar to the one-page questionnaire.
We assessed the intermethod reliability of the summary physical activity variables and individual activities as measured by the one-page questionnaire compared with the detailed interview by the Pearson partial correlation coefficient after adjustment for age (in 5-year categories) and sex. Because most of the continuous measures were positively skewed, we log-transformed all continuous variables except BMI.
Table 1 gives the distribution of demographic and lifestyle characteristics of the measurement study participants and the full VITAL cohort. Table 1 also provides summary physical activity measures stratified by demographic characteristics. According to the one-page questionnaire, 92% of the measurement-study participants reported some type of regular recreational physical activity during the preceding 10 years, compared with 85% of participants in the full cohort. In analyses limited to cohort members from the sampling area for the measurement study (King County), reported prevalence of any activity and median MET-hours per week were similar to values for the measurement-study participants (data not shown). Among those who reported any activity, measurement-study participants tended to report similar or higher MET-hours per week than those in the full cohort, and this was relatively consistent across groups defined by demographic characteristics. Both the proportion reporting some recreational activity on the one-page questionnaire and the median MET-hours per week among those who were active were generally higher in men, those with a college degree, never or former smokers, and individuals with a BMI <25 kg/m2. These differences were fairly consistent among both the measurement-study participants and the full cohort, and for the one-page questionnaire and comparison interview (data not shown).
Table 2 gives comparisons of physical activity participation and MET-hours per week by intensity level and type of activity between the one-page questionnaire and the detailed comparison interview. The proportion of individuals reporting some regular activity and the median MET-hours per week among those who were active were consistently higher as estimated by the interview compared with the one-page questionnaire. Among active men and women, the median MET-hours per week was approximately twice as high from the comparison interview than from the one-page questionnaire (30.4 versus 16.5 for men; 16.6 versus 7.1 for women). Results were similar for other categories of activity by intensity level, except for low-intensity activities among men. For individual activities, reported activity was similar on both measurement instruments for yoga and weight lifting among women. However, approximately twice as many participants reported running/jogging or swimming in the interview than on the one-page questionnaire.
Table 3 gives correlations for MET-hours per week between the one-page questionnaire and the comparison interview. The age- and sex-adjusted Pearson partial correlation coefficient for total recreational activity between the 2 instruments was 0.68. Correlations were stronger for moderate- and high-intensity activities (r = 0.58) and high-intensity activities (r = 0.52) than for moderate- or low-intensity activities alone (r = 0.27 and 0.44, respectively). The correlation coefficients were higher (r = 0.75 to 0.76) for running/jogging, yoga, weight lifting, and swimming than for walking (r = 0.55).
To determine whether correlations coefficients for recreational activity between the one-page questionnaire and the detailed interview differed by other characteristics, we calculated age- and sex-adjusted Pearson partial correlations stratified by age, BMI, and education (Table 4). Although there were no clear patterns by age or BMI among men and women, correlations were stronger among those with more education.
We also estimated the correlation between average MET-hours over the previous 10 years and BMI, which is an indirect physiologic correlate of physical activity (Table 5). In the full cohort, BMI was inversely correlated with physical activity as assessed by the one-page questionnaire (r = −0.22). This association was somewhat stronger among women (r = −0.26) than among men (r = −0.18). Among the measurement-study participants, the correlation coefficient for total MET-hours and BMI from the comparison interview and the one-page questionnaire were similar (r = −0.25, and −0.27, respectively). Correlations were stronger for high-intensity activities than for moderate- or low-intensity activities.
We evaluated the measurement characteristics of a one-page questionnaire of 10-year recreational activity compared with a detailed interview. Participants reported higher regular activity levels in the detailed interview than in the one-page questionnaire. The age- and sex-adjusted correlation of MET-hours per week from all activities between the 2 instruments was 0.68. MET-hours for high-intensity activities were more strongly correlated between the 2 instruments than moderate- or low-intensity activities alone. The strength of the correlation between MET-hours per week estimated from the 2 instruments varied by age and BMI, and was consistently stronger for both men and women with more education. Average energy expenditure was inversely correlated with BMI (r = −0.22), providing moderate support for an association, considering that multiple factors affect BMI.
Strengths of the study include development of a simple self-administered questionnaire on recreational physical activity and its comparison to a detailed interview, as well as to a physiological correlate of activity, BMI. There are also limitations to our study. We limited eligibility to individuals living in King County and we oversampled those with high supplement intakes. Consequently, measurement-study participants were more educated, were less likely to smoke and be overweight, consumed higher doses of supplements, and reported higher levels of activity than those in the full cohort.9 Furthermore, cohort members had volunteered for a study of supplement use and cancer, and are likely to be more health-conscious than the general population. If health-conscious individuals are able to report their recreational physical activity more accurately, the correlation coefficients could be overestimated.
Another potential limitation of the one-page questionnaire is the inclusion of recreation physical activities only. However, occupational activity is generally low among older adults. In our measurement-study population, only 19% of men and 11% of women worked at a strenuous- or moderate-activity job for 5 years or more in the past 10 years. Nonrecreational activities such as household and gardening activities were also not included because they are often of low-intensity and have not shown strong associations with disease. In addition, the format of the questionnaire (ie, lists of moderate and strenuous rather than separate items for each) could have led to some misclassification. Notably, participants could have been more likely to miss activities included in a checklist compared with those listed alone on the one-page questionnaire, as evidenced by the lower proportion of participants reporting such activities as running/jogging (listed in a checklist) than weight lifting (presented as a single item). Nonetheless, the correlations between the one-page questionnaire and the interview were higher for specific moderate- and high-intensity activities such as jogging/running and swimming (included as items in a single question) than for walking. Thus, it is not clear if the questionnaire would be improved if we had separate questions on high- and moderate-intensity activities. Finally, because weight was self-reported, it is possible that participants underreported their weight (and consequently BMI). However, self-reported weight has been found to be highly valid in other studies.14
There are no validation measures of 10-year physical activity. Ten years of repeated diaries would be informative but not feasible. We administered an interview in which we carefully collected details of usual physical activity in the preceding 10 years. The interview began with a set of questions about the participant’s life 10 years ago, which was aimed at providing anchors or cues that would improve recall.15 Participants could have been more likely to include activities done regularly, but only seasonally, in the interview (which asked about activities done at least 12 times per year) compared with the one-page questionnaire (which asked about activities done once per week for at least 1 year). The comparison interview recorded information for each activity separately; allowed for open-ended responses for years, frequency, and duration; and included information on intensity for most activities. There are, however, several sources of error for this instrument. Part of the apparent underreporting we observed on the one-page questionnaire might actually have been substantial overreporting of activity in the detailed interview. Several studies16,17 have found that respondents tend to overreport socially desirable behaviors in interviews compared with self-administered questionnaires. Furthermore, the interview and the one-page questionnaire were similarly correlated with our objective measure, BMI, (r = −0.25, −0.27, respectively), which calls into question whether the interview was in fact more accurate. We assume that by capturing more details, the detailed interview provides a better measure of interperson variability.
We are aware of only 417–20 prior studies that have assessed the reproducibility of long-term or lifetime physical activity questionnaires; 3 of these were recently summarized.20 All used a test–retest reliability design with repeated administration of the same questionnaire 2 weeks to 1 year apart. Briefly, the Historical Leisure Activity Questionnaire was an interviewer-administered questionnaire that assessed activity for 4 age periods (ages 14 to 21, 22 to 34, 35 to 50, and 51 to 65 years) and yielded Spearman rank coefficients of 0.69 to 0.85 for the 4 age periods.18 Chasan-Taber and colleagues20 used a modified version of the same questionnaire and observed an intraclass correlation for total lifetime recreational physical activity of 0.87. Repeat administration of a third questionnaire assessing lifetime physical activity yielded a Pearson correlation coefficient for hours per week of exercise of 0.7219; Spearman rank correlations were higher for heavy (r = 0.52) and moderate (r = 0.54) activities than for light activities (r = 0.41). Finally, an interviewer-administered, computer-assisted lifetime questionnaire obtained intraclass correlation coefficients for intraobserver reliability between 0.77 and 0.96 for 4 age periods.21 Test–retest reliability studies would be expected to have higher correlations than intermethod reliability studies such as ours as a result of correlated errors. Despite this, our correlations were similar in magnitude to those from reproducibility studies. In addition, several studies22,23 have observed reasonable reliabilities for recalled activity compared with activity that was reported up to 10 years earlier in the study, suggesting that questionnaires relying on recall can collect accurate information.
To our knowledge, no other long-term physical activity questionnaires have compared physical activity measures with BMI. However, questionnaires of short-term (current or 1-year) physical activity have used BMI as an objective comparison measure, with reported correlations in the range of −0.04 to −0.13.24,25 Thus, our questionnaire compares favorably to short-term questionnaires in its association with BMI, possibly because it could take several years of activity to better predict body composition.
In summary, our short questionnaire appeared to measure long-term physical activity with enough precision to quantitatively examine associations between physical activity and disease risk or to control for physical activity as a covariate. Although it appeared to be biased (ie, substantial underreporting of many activities compared with the detailed interview), it is precision (reflected in the correlation coefficients) rather than bias that affects risk ratios under nondifferential measurement error.15 Brief instruments that can accurately assess physical activity can be helpful in conducting studies of the effects of long-term physical activity and developing public health policies to promote physical activity.
We thank Kayla Stratton who performed some statistical analyses.
1.Hu FB, Leitzmann MF, Stampfer MJ, et al. Physical activity and television watching in relation to risk for type 2 diabetes mellitus in men. Arch Intern Med
3.Friedenreich CM. Physical activity and cancer prevention: from observational to intervention research. Cancer Epidemiol Biomarkers Prev
4.McTiernan A, Ulrich C, Slate S, et al. Physical activity and cancer etiology: associations and mechanisms. Cancer Causes Control
5.World Cancer Research Fund Panel. Food, Nutrition, and the Prevention of Cancer: A Global Perspective.
Washington, DC: American Institute for Cancer Research, 1997.
6.Kriska AM, Caspersen CJ. Introduction to a collection of physical activity questionnaires. Med Sci Sports Exerc
7.Ainsworth BE, Sternfeld B, Slattery ML, et al. Physical activity and breast cancer: evaluation of physical activity assessment methods. Cancer
8.Ainsworth BE, Haskell WL, Whitt MC, et al. Compendium of physical activities: an update of activity codes and met intensities. Med Sci Sports Exerc
9.Satia-Abouta J, Patterson RE, King IB, et al. Reliability and validity of self-report of vitamin and mineral supplement use in the vitamins and lifestyle (VITAL) study. Am J Epidemiol
10.Ainsworth BE, Haskell WL, Leon AS, et al. Compendium of physical activities: classification of energy costs of human physical activities. Med Sci Sports Exerc
11.Jacobs DR Jr, Ainsworth BE, Hartman TJ, et al. A simultaneous evaluation of 10 commonly used physical activity questionnaires. Med Sci Sports Exerc
12.Taylor HL, Jacobs DR Jr, Schucker B, et al. A questionnaire for the assessment of leisure time physical activities. J Chronic Dis
13.Richardson MT, Leon AS, Jacobs DR Jr, et al. Comprehensive evaluation of the Minnesota leisure time physical activity questionnaire. J Clin Epidemiol
14.Rimm EB, Stampfer MJ, Colditz GA, et al. Validity of self-reported waist and hip circumferences in men and women. Epidemiology
15.Armstrong BK, White E, Saracci R. Principles of Exposure Measurement in Epidemiology
. Oxford: Oxford University Press; 1994.
16.Okamoto K, Ohsuka K, Shiraishi T, et al. Comparability of epidemiological information between self- and interviewer-administered questionnaires. J Clin Epidemiol
17.Vuillemin A, Oppert JM, Guillemin F, et al. Self-administered questionnaire compared with interview to assess past-year physical activity. Med Sci Sports Exerc
18.Kriska AM, Sandler RB, Cauley JA, et al. The assessment of historical physical activity and its relation to adult bone parameters. Am J Epidemiol
19.Friedenreich CM, Courneya KS, Bryant HE. The lifetime total physical activity questionnaire: development and reliability. Med Sci Sports Exerc
20.Chasan-Taber L, Erickson JB, McBride JW, et al. Reproducibility of a self-administered lifetime physical activity questionnaire among female college alumnae. Am J Epidemiol
21.Vuillemin A, Guillemin F, Denis G, et al. A computer-assisted assessment of lifetime physical activity: reliability and validity of the Quantap software. Rev Epidemiol Sante Publique
22.Blair SN, Dowda M, Pate RR, et al. Reliability of long-term recall of participation in physical activity by middle-aged men and women. Am J Epidemiol
23.Lee MM, Whittemore AS, Lung DL. Reliability of recalled physical activity, cigarette smoking, and alcohol consumption. Ann Epidemiol
24.Salonen JT, Lakka T. Assessment of physical activity in population studies—validity and consistency of the methods in the Kuopio ischemic heart disease risk factor study. Scand J Sports Sci
25.Washburn RA, Smith KW, Goldfield SR, et al. Reliability and physiologic correlates of the Harvard alumni activity survey in a general population. J Clin Epidemiol