Background:

Men and women who exercise regularly and who are physically fit tend to be leaner than those who are sedentary and not fit. Although exercise is known to attenuate weight gain and promote weight loss, there may also be a propensity for leaner men and women to choose to exercise vigorously (self-selection). Pre-exercise body weights have been shown to account for all the weight differences between fast and slow walkers, but seem to account for only a portion of the weight differences associated with walking distances. Whether these results apply to maximum exercise performance (i.e., cardiorespiratory fitness) as well as to doses of vigorous exercise (metabolic equivalents >6) remains to be determined.

Objective:

Assess whether the cross-sectional relationships of BMI to cardiorespiratory fitness and vigorous activity are explained by BMI prior to exercising.

Methods and Procedures:

Cross-sectional study of the relationships between cardiorespiratory fitness (running speed during 10 km foot race) and vigorous physical activity (weekly running distance) to current BMI (BMI_current) and BMI at the start of running (BMI_starting) in 44,370 male and 25,252 female participants of the National Runners' Health Study.

Results:

BMI_starting accounted entirely for the association between fitness and BMI_current in both sexes, but only a quarter of the association between vigorous physical activity levels and BMI_current in men. In women, BMI_starting accounted for 58% of the association between BMI_current and vigorous activity levels.

Discussion:

Self-selection based on pre-exercise BMI accounts entirely for the association found between fitness and BMI (and possibly a portion of other health outcomes).

Introduction

Cross-sectional and prospective studies show that BMI and other measures of ponderosity are inversely related to cardiorespiratory fitness and physical activity (1,2). Fitness varies substantially among individuals due to age, genetic constitution, and health, independent of physical activity (3,4,5). It is not known whether the leanness typically associated with cardiorespiratory fitness is the direct consequence of exercise training. It could also be due to self-selection, i.e., a propensity for leaner men and women to choose to exercise vigorously.

We have previously demonstrated that self-selection accounts for most of the cross-sectional decline in BMI associated with walking intensity and significant proportions of the decline in BMI associated with walking distance (6). This finding may not apply to vigorous exercise because the impediments to exercising at moderate and vigorous intensities may differ. Walking is considered a moderately intense activity, i.e., one that increases energy expenditure by less than 6-fold over resting levels (generally 4-fold (7)) whereas running is classified as a vigorously intense activity, i.e., one that increases energy expenditure from 7- to 18-fold or more (generally 8-fold (7)).

The current report uses a large cohort of runners (from The National Runners' Health Study) to assess whether self-selection accounts for the inverse association of runners' weights with and vigorous physical activity and fitness levels observed cross-sectionally. Together with our prior results for walking, these results provide an assessment of the self-selection bias across the spectrum of moderate to vigorous exercise.

Methods and procedures

The design and methods of the National Runners' Health Study are described elsewhere (8,9,10). Briefly, recruitment took place during two phases, the first between 1991 and 1993 and the second between 1998 and 1999. A two-page questionnaire, distributed nationally at races and to running magazine subscribers, solicited information on demographics (such as age, race, education); running history (such as age when the person began running at least 12 miles/week, average weekly mileage and number of marathons run over the preceding 5 years, times for the best marathon run and for the 10 km race); weight history (greatest and current weight, weight when the person started running, least weight as a runner, body circumferences of the chest, waist, and hips); diet (vegetarianism, and the current weekly intakes of alcohol, red meat, fish, fruit; vitamin C, vitamin E, and aspirin); current and past cigarette use; history of heart attacks and cancer; and medications used for blood pressure, thyroid, cholesterol or diabetes. Physical activity histories prior to running 12 or more miles/week were not solicited. Running distances were reported in miles run per week, body circumferences in inches, and body weights in pounds. These values were converted to kilometers, centimeters, and kilograms for this report.

Although other leisure-time physical activities were not recorded for many in this cohort, data from runners recruited after 1998 (when the question was introduced) show that running represented (s.d.) 91.5 19.1% and 85.2 24.0% of all vigorously intense activity in men and women, respectively, and 73.5 23.7% and 69.4 25.7% of total leisure-time physical activity. Self-reported weekly running distances were found to be strongly correlated with distances reported in repeated questionnaires (r = 0.89 in 110 men and 40 women). There were strong correlations between repeated questionnaires for self-reported and clinically measured height (r = 0.96) and weight (r = 0.96), and for self-reported running distance vs. self-reported BMI and waist circumference in cross-sectional analyses (9,10). Self-reported and clinically measured waist circumferences were moderately correlated (r = 0.68). All participants signed a statement of informed consent, and the review board of the University of California at Berkeley approved the study protocol.

Cardiorespiratory fitness in this report is defined as the speed (m/s) of the participant's best 10-km race during the previous 5 years. Published data support the use of running performance for estimating maximal oxygen consumption (VO_2max) (11,12,13).

Statistical analyses

Multiple linear regression was used to adjust for adiposity, physical activity, and fitness using both age and age² as covariates. Age squared (age²) was included as a covariate to accommodate some degree of nonlinearity in the relationship between age and BMI. The effects of self-selection were estimated by the same three approaches used previously in our study of self-selection in walkers (6). Specifically, fitness and physical activity are compared to the current BMI (BMI_current) and to the BMI when participants first began running 12 or more miles/week (BMI_starting), as well as to current and starting waist, hip and chest circumferences, and bra-cup sizes. To assess the proportion of the BMI differences between deciles of activity or fitness attributable to self-selection we: (i) divided the sample into deciles of fitness or physical activity, (ii) calculated the decline in BMI_current at each decile relative to the lowest decile of fitness or activity, and (iii) compared these declines to the corresponding differences in BMI_starting. Self-selection was calculated as a proportion of the mean differences in BMI_current represented by the mean differences in starting BMI. These analyses were further confirmed by multiple regression analyses (6) and percentile regression analyses (6). The statistical software for calculating the polynomial regression was JMP Version 5.1 (SAS Institute, Cary, NC).

Results

There were 44,370 men and 25,252 women who provided complete information on age, weekly running distance, 10-km race performance, and weight when they began running 12 or more miles/week, who did not use thyroid or diabetic medications, who were nonsmokers, and who were not strict vegetarians. They were generally middle-aged (mean s.d., men: 44.47 10.58, women: 38.51 9.61 years), and lean (BMI men: 23.96 2.68; women: 21.37 2.42 kg/m²). The men and women had run an average of 39.79 22.20 and 37.64 20.70 km/week, had run 12 or more miles/week for 12.7 8.1 and 9.7 6.6 years, and had run a 10-km race at an average speed of 3.91 0.56 and 3.42 0.54 m/s, respectively. In addition, 37,850 men and 19,959 women reported their current and past waist circumferences, 16,413 men and 19,077 women reported current and past hip circumferences, 30,171 men and 20,035 women reported current and past chest circumferences, and 21,181 women reported their bra-cup sizes.

Analysis of mean differences

Figure 1 presents the age-adjusted histograms of the mean difference in BMI_current and current waist circumference between the least-fit men (1st decile) and men of the 2nd through 10th deciles of fitness. The differences in average BMI_current between the lowest and higher fitness categories increased progressively with deciles of fitness. However, the differences in average BMI_starting between the least fit and higher fitness categories also increased progressively with fitness and accounted for >97% of the differences in BMI_current between the least fit and fitter men. Differences in average BMI_current between the least active and higher activity categories also increased progressively with physical activity (running distance), however the portions of these differences accounted for by differences in average BMI_starting were small. Similar results were obtained for waist circumference (a measure of intra-abdominal fat).

Figure 1.

Self-selection in men. Values below bars represent the proportions accounted for by self-selection, which were estimated: (i) by dividing the sample into deciles of fitness or physical activity, (ii) by calculating the decrease in BMI_current at each decile relative to the lowest decile of fitness or activity, and (iii) by comparing these decreases to the corresponding differences in starting BMI (specifically their recollection of adiposity when they first began running 12 or more miles/week). Self-selection was calculated as the proportion of the mean reduction in BMI_current represented by the mean decrease in starting BMI values. Negative heights mean fitter men were leaner. All variables were age-adjusted.

Full figure and legend (34K)

The corresponding analyses of women (Figure 2) shows that differences in BMI_current, current hip circumference, and current bra-cup size between the least fit and fitter women also increased progressively with fitness but that these differences were all accounted for by differences in BMI_starting and starting body size. Women who ran greater weekly distances were also leaner than less active women. Less than a third of the declines in BMI_current and body size with physical activity can be ascribed to BMI_starting and starting body size, except for BMI_current in the 7th through 10th deciles, and current hip circumference in the 9th and 10th deciles.

Figure 2.

Self-selection in women. Histogram of the mean difference in current BMI, hip circumference, and bra-cup size between the least-fit women (1st decile) and women of the 2nd through 10th deciles of fitness (left) and between the least active women (1st decile) and women of the 2nd through 9th decile of physical activity (right). Negative heights mean fitter women were leaner. Adiposity, physical fitness, and activity are all age-adjusted.

Full figure and legend (30K)

We also used regression analyses to estimate the proportion of the regression slopes relating BMI_current and current body sizes to fitness, and those that were attributable to BMI_starting and starting body sizes (analyses not displayed). On average, starting values accounted for 93, 95.0, 92.7, and 94.1% of the regression slope of BMI_current, and current waist, hip, and chest circumferences vs. cardiorespiratory fitness, respectively, in men, and 99.2, 97.9, 99.99, 97.8, and 95.7% of the slopes relating BMI_current, and current waist circumference, hip circumference, chest circumference, and bra-cup size vs. cardiorespiratory fitness, respectively, in women. In contrast, starting values accounted for only 24.0, 14.5, 28.1, and 27.4% of the regression slope of BMI_current, and current waist, hip, and chest circumferences, vs. physical activity level in men. The percentages were slightly higher for women, i.e., starting values accounted for 54.5% of the decline in BMI_current with physical activity level, 48.0% of the decline in current waist, 52.1% of the decline in current hip, and 43.6% of the decline in current chest circumference, and 25.3% of the decline in current bra-cup size/km/week run.

Percentile regression

Figure 3 displays relationships of the 5th, 25th 50th, 75th, and 95th percentiles of BMI_current to fitness. The relationships are convex, and the tangent slopes for BMI_current vs. fitness become progressively greater (more negative) at higher BMI percentiles in both men and women. In both sexes the tangent slopes for BMI_current vs. fitness are almost entirely accounted for by the tangent slopes for BMI_starting vs. fitness. As to physical activity levels (Figure 4), the proportions of the tangent slopes for BMI_current vs. km/week run and those represented by BMI_starting vs. km/week run are generally small, being least among the least active and achieving somewhat greater proportions with increasing activity among the leanest subjects. All of the association between physical activity and BMI_current is attributable to BMI_starting among the leanest men and women who are also the most active; however, at these distances the incremental decline in BMI per km/week is small. When averaged over the distributions of fitness and physical activity, taking into account differences between BMI percentiles, we attribute 93.9% of the decline in BMI_current for men and 95.5% for women,with physical fitness, to the differences in BMI before starting to run; and 26.4% of the decline in BMI_current for men and 58% for women,with physical activity, to their starting BMI.

Figure 3.

Percentile regression. The dashed lines display the relationships of the 5th, 25th, 50th, 75th, and 90th percentiles of the current (BMI_current) and starting BMI (BMI_starting) to cardiorespiratory fitness (m/s during 10-km race). The line segments are the tangents to these percentile curves at 3.2, 3.9, and 4.6 m/s in men and 2.4, 3.2, and 4.0 m/s in women. The slope of each tangent line segment is presented proximal to the line segments. All variables are age-adjusted.

Full figure and legend (33K)

Figure 4.

Percentile regression. The dashed lines display the relationships of the 5th, 25th, 50th, 75th, and 90th percentiles of the current (BMI_current) and starting BMI (BMI_starting) to physical activity (km/week). The line segments are the tangents to these percentile curves at 0, 35 and 70 km/week. The slope of each tangent line segment is presented proximal to the line segments. All variables are age-adjusted.

Full figure and legend (33K)

Discussion

We have shown that self-selection (or the tendency for initially leaner men and women to be also fitter initially or become faster (i.e., fitter) runners), accounts for the association between cardiorespiratory fitness and current leanness in a large cross-sectional sample of runners. This was demonstrated in both men and women and with respect to a variety of adiposity measures that include fat deposits that are characteristically masculine or feminine. The ability to run fast may have been a pretraining characteristic of the leaner runner. The findings suggest that the leanness of the fittest runners may not reflect the effect of fitness on BMI, but rather the effect of BMI on cardiorespiratory fitness and speed. Our analyses (which proceed from the simplest comparisons of the differences between deciles, regression analyses recognizing the nonlinearity of the relationships between adiposity and fitness and physical activity, and percentile regression recognizing the dependence of the relationships on the percentiles of BMI), were remarkably consistent in the average percentage attributable to self-selection. However, each of the approaches revealed in somewhat greater detail the dependence of the bias on the level of fitness (quadratic regression) and BMI (percentile regression).

We have previously demonstrated that the decline in body weight and size with walking intensity was entirely accounted for by body weight and size when starting to exercise, but that only a portion of the decline with walking distance was accounted for by their pre-exercise values (6). Also as observed in these runners, the proportion of the decline in BMI_current with walking distance accounted for by BMI_starting was greater in men (58%) than women (31%). Thus whereas the population of runners studied here represent a select, active portion of the population, when considered together with the walkers the results span the range of exercise intensity from moderately intense (3–6 metabolic equivalents) to vigorously intense (>6 metabolic equivalents), and reveal consistent differences between exercise dose and intensity. Both studies show that the decline in body weights per km/week of were more reflective of starting weights in women than men.

Self-selection and cardiorespiratory fitness

The apparent health benefits of physical activity rely strongly on studies of cardiorespiratory fitness (14). Yet analyses presented in this paper suggest there is greater self-selection when body weight and size are compared to cardiorespiratory fitness than to physical activity. Genes contribute to differences in cardiorespiratory fitness independent of physical activity. Rats selectively bred for treadmill endurance achieve a 58% improvement in mean distance run until exhaustion after one generation (5) and a 70% improvement after three generations (4). Data from twins suggest that as much as 93% of maximum aerobic power in unconditioned individuals may be genetically determined (3), and data from young adults suggest genetics account for approximately 70% of total work performed during a 90-min maximal ergocycle (15). Cardiac output and stroke volume, factors that contribute to maximum aerobic capacity, also exhibit significant inheritance in sedentary individuals and for changes in responses to endurance training (16). The HERITAGE study reported a maximal heritability estimate of 47% for increases in maximum aerobic consumption in family members after 20 weeks of training (17). A high innate initial VO_2max (not measured) may also be an important determinant for becoming a runner in the first place, the weekly distance run, and time taken for the 10-km race.

Self-selection and physical activity

The current analyses show that starting weight also affects the quantity of physical activity performed weekly, and that for 26% (of men) and 58% (of women), the association between physical activity and BMI may be due to initially leaner men and women running longer distances. These results are consistent with the observations by others that body weight is a barrier to being physically active (18) and that body weight predicts inactivity in prospective epidemiological studies (19). Weight differences between active and sedentary older women have been shown to trace back to their weights during young adulthood (20). Self-selection would explain why the relationship between adiposity and physical activity is more easily documented in cross-sectional observational studies than training studies. Specifically, self-selection augments cross-sectional associations but not longitudinal associations of change.

Summary

We have demonstrated that self-selection accounts for all of the association between fitness and BMI and a nontrivial portion of the association between physical activity and BMI in men and women who engage in at least some vigorous activity. Current estimates of the dose of physical activity required to maintain healthy weight (21) are reflective in part of the effect of body weight on the propensity to be physically active, in addition to the effect of physical activity on the maintenance of healthy weight.

DISCLOSURE

The authors declared no conflict of interest.

References

REFERENCES

1. Wong SL, Katzmarzyk P, Nichaman MZ et al. Cardiorespiratory fitness is associated with lower abdominal fat independent of body mass index. Med Sci Sports Exerc 2004;36:286–291. | Article | PubMed | ISI |
2. DiPietro L, Kohl HW 3rd,Barlow CE, Blair SN. Improvements in cardiorespiratory fitness attenuate age-related weight gain in healthy men and women: the Aerobics Center Longitudinal Study. Int J Obes Relat Metab Disord 1998;22:55–62. | ChemPort |
3. Klissouras V. Heritability of adaptive variation. J Appl Physiol 1971;31:338–344. | PubMed | ChemPort |
4. Koch LG, Meredith TA, Fraker TD, Metting PJ, Britton SL. Heritability of treadmill running endurance in rats. Am J Physiol 1998;275:R1455–R1460. | PubMed | ChemPort |
5. Troxell ML, Britton SL, Koch LG. Selected Contribution: variation and heritability for the adaptational response to exercise in genetically heterogeneous rats. J Appl Physiol 2003;94:1674–1681. | PubMed |
6. Williams PT. Self-selection contributes significantly to the lower adiposity of faster, longer-distanced, male and female walkers. Int J Obes (Lond) 2007;31:652–662. | Article | PubMed | ChemPort |
7. Ainsworth BE, Haskell WL, Leon AS et al. Compendium of physical activities: classification of energy costs of human physical activities. Med Sci Sports Exerc 1993;25:71–80. | Article | PubMed | ISI | ChemPort |
8. Williams PT. Relationship of distance run per week to coronary heart disease risk factors in 8283 male runners. The National Runners' Health Study. Arch Intern Med 1997;157:191–198. | Article | PubMed | ISI | ChemPort |
9. Williams PT, Satariano WA. Relationships of age and weekly running distance to BMI and circumferences in 41, 582 physically active women. Obes Res 2005;13:1370–1380. | PubMed |
10. Williams PT, Pate RR. Cross-sectional relationships of exercise and age to adiposity in 60, 617 male runners. Med Sci Sports Exerc 2005;37:1329–1337. | Article | PubMed |
11. Balke B, Ware RW. An experimental study of physical fitness of Air Force personnel. US Armed Forces Med J 1959;10:875–888.
12. Hellerstein HK. Limitations of marathon running in the rehabilitation of coronary patients: anatomic and physiologic determinants. Ann NY Acad Sci 1977;301:484–494. | Article | PubMed | ChemPort |
13. Cooper KH. A means of assessing maximal oxygen intake: correlation between field and treadmill testing. JAMA 1968;203:201–204. | Article | PubMed | ChemPort |
14. Pate RR, Pratt M, Blair SN et al. Physical activity and public health. A recommendation from the Centers for Disease Control and Prevention and the American College of Sports Medicine. JAMA 1995;273:402–407. | Article | PubMed | ISI | ChemPort |
15. Bouchard C, Lesage R, Lortie G et al. Aerobic performance in brothers, dizygotic and monozygotic twins. Med Sci Sports Exerc 1986;18:639–646. | PubMed | ChemPort |
16. An P, Rice T, Gagnon J et al. Familial aggregation of stroke volume and cardiac output during submaximal exercise: the HERITAGE Family Study. Int J Sports Med 2000;21:566–572. | Article | PubMed | ChemPort |
17. Bouchard C, An P, Rice T et al. Familial aggregation of VO2max response to exercise training: results from the Heritage family study. J Appl Physiol 1999;87:1003–1008. | PubMed | ChemPort |
18. Ball K, Crawford D, Owen N. Too fat to exercise? Obesity as a barrier to physical activity. Aust NZ J Public Health 2000;24:331–333. | ChemPort |
19. Petersen L, Schnohr P, Sørensen TI. Longitudinal study of the long-term relation between physical activity and obesity in adults. Int J Obes Relat Metab Disord 2004;28:105–112. | Article | PubMed | ChemPort |
20. Voorrips LE, Meijers JHH, Sol P, Seidell JC, van Staveren WA. History of body weight and physical activity of elderly women differing in current physical activity. Int J Obes 1992;16:199–205. | ChemPort |
21. Institute of Medicine. Dietary Reference Intakes for Energy, Carbohydrate, Fiber, Fat, Fatty Acids, Cholesterol, Protein, and Amino Acids (Macronutrients). The National Academies Press: Washington DC, 2005936pp.

Acknowledgments

Dr Williams was responsible for obtaining funding, overseeing the study implementation, statistical analyses, and manuscript preparation. This study was supported in part by grants HL-45652, HL-072110 and DK066738 from the National Heart Lung and Blood Institute, and was conducted at the Lawrence Berkeley Laboratory (Department of Energy DE-AC03-76SF00098 to the University of California).