Article Text
Abstract
Objective: To assess the association of weekly walking distance to body weight and waist circumference in elderly (age ⩾75 years), senior (55⩽ age <75 years), middle-aged (35⩽ age <55 years), and younger men (18⩽ age <35 years old).
Design: Cross-sectional analyses of baseline questionnaires from 7082 male participants of the National Walkers’ Health Study.
Results: Standard regression analyses showed that body mass index (BMI) was inversely and significantly associated with walking distance (kg/m2 per km/week) in elderly (slope (SE): −0.032 (0.008)), senior (−0.045 (0.005)) and middle-aged men (−0.037 (0.007)), as were their waist circumferences (−0.090 (0.025), −0.122 (0.012) and −0.091 (0.015) cm per km/week, respectively), and that these slopes remained significant when adjusted statistically for reported weekly servings of meat, fish, fruit and alcohol. However, percentile regression analyses showed that the declines in BMI per km/week walked were greater at the higher than the lower percentiles of the BMI distribution. In men ⩾74 years old the decline per km walked was 4.9-fold greater among the heaviest men (that is, 90th BMI percentile; −0.076 kg/m2 per km/week) than among the leanest men (that is, 10th BMI percentile; −0.015 kg/m2 per km/week). The differences in slope at the 90th compared to the 10th BMI percentile were 5.4-fold among men 55–74 years old and sixfold among men 35–54 years old. Per km/week walked, the declines at the 90th percentile of waist circumference were also greater than at its 10th percentile, and intermediate for percentiles in between. Whereas standard regression analyses suggest that the average declines in BMI per km/week walked reported here are consistent with those reported previously per km/week run in male runners 35–54 years old (−0.036 (0.001) kg/m2 per km/week) and ⩾50 years old (−0.038 (0.001) kg/m2 per km/week), percentile regression analyses showed that when adjusted to the leaner body weights of the runners the declines per km walked were between 49% and 59% less for walkers than runners.
Conclusions: Declines in BMI and waist circumferences with walking distance depend upon the percentile of the BMI distribution, with the decline per km walked being significantly greater among heavier men.
Statistics from Altmetric.com
Men who are physically active during leisure are less likely to be obese.1 2 Walking is among the most popular recreational activities,3–5 and is specifically endorsed to meet current health recommendations by government and non-government organisations.6–8 Elderly men are more likely to choose walking for exercise than men more junior.9
The dose-response relation between walking distance and adiposity is not well established. Some cross-sectional studies,10 particularly those employing pedometers rather than survey instruments,11–13 demonstrate an association between walking distance and measures of adiposity. Other study designs provide mixed findings on whether walking is sufficient to promote weight loss or prevent weight gains over time.14–17 Substantial walking effort may be required to maintain healthy weight—for example, 45–60 minutes daily of brisk walking has been suggested for preventing meaningful weight gain in men during middle age.18 19
The dose-response relation may also depend upon age, with some studies suggesting that differences in body mass index (BMI) and body fat mass between physically active and sedentary men increase from 18–64 years old, then diminish somewhat thereafter.20 Different associations in older and younger men could relate to age-related differences in fat accumulation and loss of lean body mass. Fat-free mass is generally stable until age 60 and decreases thereafter.21 22 Fat mass increases during middle age and possibly thereafter until age 75.21
The current study examines the associations of self-reported weekly walking distance to percentiles BMI and waist circumferences in a large sample of primarily older male walkers. Previous cross-sectional studies have primarily described the dose-response relations of walking to adiposity by means, regression slopes and proportions that exceed threshold levels (that is, ⩾25 kg/m2 or ⩾30 kg/m2 23). We have shown in female walkers24 and in male and female runners25 26 that the dose of exercise has a greater apparent effect on the 90th BMI percentile than the 10th BMI percentile, and a graded apparent effect at intermediate percentiles.
METHODS
A two-page questionnaire was mailed to walkers identified through a walking magazine subscriber list.25 Approximately 8% of the 575 000 subscribers solicited elected to join the National Walkers’ Health Study. Our goal was to obtain a sufficiently large cohort for a prospective epidemiological study of health in walkers rather than a comprehensive survey of these magazine subscribers, thus recruitment among subscribers ceased once more than 50 000 questionnaires were received (including multiple surveys from the same individuals). Subscribers were primarily women who chose Walking Magazine through stamp-sheet sweepstakes and thus included an ethnically and socially diverse readership. The study protocol was reviewed by the University of California Berkeley Committee for the Protection of Human Subjects, and all subjects provided a signed statement of informed consent.
Each walker submitted a two-page questionnaire that solicited information on demographics (age, race, education), walking history (age when began walking at least 12 miles per week, average weekly mileage, walking frequency, longest walk, usual walking speed), weight history (greatest and current weight, weight when started walking, least weight as a walker, 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, previous history of heart attacks and cancer, and medications for blood pressure, thyroid, cholesterol or diabetes. Walking distances were reported in miles per week, body circumferences in inches and body weights in pounds. These values were converted to kilometres, centimetres and kilograms for this report.
BMI was calculated as self-reported weight in kilograms divided by self-reported height in meters squared. Self-reported waist circumferences were in response to the question “Please provide, to the best of your ability, your body circumference in inches” without further instruction. The relation of waist circumference with walking distance may be weakened by different perceptions of where the circumference lies. However, unless the perceived location varies systematically in relation to distance, this subjectivity is unlikely to produce the associations reported in the tables and figures. The circumference dimensions, rather than their ratios, are reported because waist circumference has been shown to be a better indicator of intra-abdominal fat.27 We have observed strong correlations between self-reported and clinically measured heights and weights (unpublished correlation in 110 men were r = 0.96 for both). We observed that self-reported waist circumferences were somewhat less precise, as indicated by their correlations with reported circumferences on a second questionnaire (r = 0.84) and with their clinical measurements (r = 0.68).
Statistical analyses
Statistics are presented as mean (SE) or regression slopes (SE) unless otherwise noted. The associations of body weight and size to walking distance were assessed graphically by determining the average adiposity within predetermined distance categories (<2 km/week, 2–14.9 km/week, 15–29.9 km/week, 30–44.9 km/week and ⩾45 km/week), separately by age group. In addition, simple regression slopes were determined by standard linear regression with BMI or waist circumference as the dependent variables and walking distance (km/week) as the independent variable. Adjusted slopes were computed with weekly intakes of meat, fish, fruit and alcohol included as covariates.
Our approach24–30 for estimating the slope for the kth percentile of Y (dependent variable) versus X (independent variable) involves partitioning the independent variable into deciles and determining the percentiles of dependent variable within each partition. Simple linear regression is then used to calculate the slope of the kth percentile of Y by X. Robust estimates of the regression coefficients, standard errors and significance levels are obtained by bootstrap resampling.31
RESULTS
Of the 8539 men who provided complete information on age and weekly walking distance, we excluded 465 men for their thyroid medication use, 591 men for their diabetes medication use, 379 men for reporting that they smoked cigarettes currently and 85 men for their strict vegetarian diets. Of the remaining 7398 men, 7082 (95.7%) provided complete heights and weights so that BMI could be calculated and 6015 men (81.3%) reported waist circumferences. Table 1 presents the characteristic of the sample by walking distance. It shows that longer-distanced walkers were slightly younger and consumed more fish, fruit and alcohol, and ate less meat than those who walked less (p<0.0001).
Figure 1 (upper panel) plots the average body mass index by weekly walking distance stratified by age. The graphs demonstrate consistent declines in BMI with increasing weekly distance in men aged 35 and older. The corresponding regression slopes appear in table 2, along with the slopes for waist circumference versus distance walked. In agreement with figure 1, the slopes were significant in men aged 35 and older but not in younger men, and remained significant when adjusted for reported weekly servings of fruit, fish, meat and alcohol. When age groups are more broadly defined as senior men (aged 55–74 years) and middle-aged men (aged 35–54 years), the slopes were −0.045 (0.005) and −0.037 (0.007) kg/m2 per km/week, respectively, for BMI and −0.122 (0.012) and −0.091 (0.015) cm per km/week, respectively, for waist circumference.
The lower panel presents the associations of weekly walking distance to the 10th, 50th (median) and 90th percentile of BMI. The decline in BMI with distance walked was weakly discernible at the 10th BMI percentile (leanest walkers), apparent at the 50th percentile and most pronounced at the 90th percentile (heaviest walkers) in men 35 years and older.
Figure 2 provides detailed analyses of the associations of weekly walking distance to the percentiles of BMI and waist circumference. The vertical axis gives the regression slope (BMI or waist circumference vs weekly walking distance). The horizontal axis gives the percentiles of BMI (upper panel) or waist distribution (lower panel). Curves are presented for elders (men aged 75 and older), senior (55–74 years old) and middle-aged male walkers (35–54 years old). The percentiles whose slopes are statistically different from zero (β≠0 at p<0.05) are designated by the solid portions of the bars at the bottom of each graph. For example, in seniors the decrease in BMI per km/week walked was −0.018 at the 10th BMI percentile, −0.025 at the 25th percentile, −0.036 at the 50th percentile (median), −0.051 at the 75th percentile and −0.099 at the 90th percentile, and the regression slopes were all significant for the 17th through the 95th sample percentiles, inclusive. The slope (decline in BMI per km walked) became progressively more negative at higher percentiles of the sample distribution, particularly above the median. The regression slopes of the elderly, senior and middle-aged men did not differ significantly from each other at any percentile (analyses not displayed). Compared to the decline at the 10th BMI percentile, the decline at the 90th BMI percentile was 4.9-fold greater in men aged 75 and older, 5.4-fold greater in men aged 55–74 and sixfold greater in men 35–54 years old.
The bottom panel of figure 2 displays the corresponding regression slopes for waist circumferences versus walking distance. In all age groups, the decrease per km/week walked became progressively greater for higher percentiles of the waist circumference. Statistical significance was generally achieved above the median waist circumference and their slopes did not differ significantly between age groups at any percentile. The oscillations arise from the discreteness of self-reported waist circumferences (generally rounded to the nearest inch).
DISCUSSION
This report shows that the walkers’ BMIs and waist circumferences were inversely and significantly associated with weekly walking distance in elderly, senior and middle-aged men. In men ⩾35 years old, there were no statistically significant differences in the slopes between age groups. It also shows that a simple regression slope or mathematical function does not adequately describe the decline in weight and body size with walking distance. The greatest decline in BMIs and waist circumferences per km/week walked was for the highest percentiles of body weight and size, suggesting that walking may most favourably influence weight among those having the greatest weight-related health risks. We also found waist circumference, an indicator of abdominal visceral fat,32 also declined in association with weekly walking distance in all five age classes of men over the age 35 years old. Preventing gains in abdominal visceral fat may be particularly important in avoiding the health complications associated with metabolic syndrome, such as hypertension, diabetes and cardiovascular disease.33
We found no significant relation between walking distance and either BMI or waist circumferences in 18–35 year old men. The sample size for this group was admittedly small, and thus provided limited statistical power to detect an association, however, the non-association in younger walkers is also consistent with the non-association we previously reported for male runners 18–25 years old.25 They are also consistent with cross-sectional data suggesting smaller differences in BMI between sedentary and physically active younger men in relation to older men.20
In this paper, we reported that the decline in BMI associated with walking distance was greater at the higher (for example, 90th) than lower (for example, 10th) percentiles of the BMI distribution. This was observed despite the purported tendency for overweight individuals to over-report their physical activity,34 which we would expect to weaken the slope. Others have also observed that the attenuation of age-related weight gain by exercise is greater in overweight than in normal weight men.35 The dependence of the slope on the percentile of the BMI distribution has additional implications. In particular, it suggests that the traditional regression slope: (1) is relevant to only a small proportion of the population distribution; (2) substantially overestimates the relation in the majority of lean men; and (3) substantially underestimates the relation in men of greatest clinical interest—those who are most overweight and at greatest morbidity and mortality risk. For example, the traditional regression slope for BMI vs distance in 55–74 year old walkers has a 95% confidence interval from −0.035 to −0.055 kg/m2 per km/week. Comparing this interval with the curve in the upper panel of figure 2 for these men shows that this confidence interval includes the slopes between the 60th and 74th percentile of the BMI distribution, thereby overestimating the relation in the lower 59th percentile of the sample and underestimating the relation in the upper 25th percentile. Different slopes for different percentiles of BMI and waist circumferences also have important implications for the statistical analyses of epidemiological data. For example, statistical adjustment for BMI presumes that the functional relations of physical activity to adiposity are parallel for all percentiles of the population distribution, which figure 2 shows is not true. Whereas standard regression analyses suggest that the average declines in BMI per km/week walked reported here are consistent with those reported previously per km/week run in male runners 35–54 (−0.036 (0.001) kg/m2 per km/week) and ⩾50 years old (−0.038 (0.001) kg/m2 per km/week),25 percentile regression analyses showed that when adjusted to the leaner body weights of the runners the declines per km walked were between 49% and 59% less for walkers than runners (analyses not displayed).
Caveats and limitations
Our data are cross-sectional and observational, so it is not possible to separate the effects of self-selection from exercise-induced weight loss. As with other large epidemiological prospective studies (for example, the Nurses' Health Study,36 Women’s Health Initiative Observational Study37), our results necessarily rely on self-reported walking distance and weight measures, pedometers and direct clinical measurements being impractical in large epidemiological cohorts. However, the correlations we observed between self-reported body weight and height (see Methods) exactly match those reported for the Nurses' Health Study. These limitations not withstanding, our findings are relevant to lifestyle choices in older men, who are more likely to choose walking for physical activity than younger men.9
REFERENCES
Footnotes
Competing interests: None.
Funding: Supported in part by grant grants HL-45652 and HL-72110 from the National Heart Lung and Blood Institute, and DK-066738 from the Institute of Diabetes and Digestive and Kidney Disease, and was conducted at the Lawrence Berkeley Laboratory (Department of Energy DE-AC03-76SF00098 to the University of California).