Our three primary findings are: (1) even among the most vigorously active populations, age-related weight gain occurs through middle-age; (2) changes in vigorous activity are associated with changes in weight in a dose-dependent manner; (3) changes in vigorous activity are associated with significantly greater changes in weight in men than in women. Prior observational studies of physical activity and adiposity have been criticized for the low prevalence of higher intensity physical activity, the measurement error associated with low-intensity activity and the inappropriate time frame of the assessment [14
]. The men and women studied here nearly all engaged in running, which is a well-quantified activity that had been sustained over many years ().
Our data lend essential support for the hypothesis that vigorous exercise promotes leanness. As our analyses are based on changing levels of exercise, the associations are unlikely to arise from lean men and women choosing to run (albeit changes in weight could affect exercise participation). Intervention studies would provide stronger evidence for causal relationship between change in weight and change in adiposity than the prospective observations we report. However, it is unlikely that any intervention studies will include the sample size (nearly 13 000 vigorously active men and women), duration (3.2 and 2.6 years of follow-up in men and women, respectively) or amount of activity (running approximately 40 km/week) reported here.
In formulating public health recommendations, there has been little discussion of the inevitability of age-related weight gain, or acknowledgement that gaining weight may be naturally associated with the aging process. Weight gain has been primarily treated as a behavioral inadequacy requiring behavioral interventions. Yet, even among runners who run 64 or more km/week, there is statistically significant weight gain over time. The caloric expenditures of these runners greatly exceed the 3.5–5 h/week of moderate intensity exercise (e.g. brisk walking) recommended by the American College of Sports Medicine to facilitate the maintenance of long-term weight loss [16
]. They also exceed other recommendations for achieving weight maintenance (e.g. 35 min of vigorous activity per day [17
], 45–60 min [18
] or 60 or 80 min of moderate intensity activity [6
], or 1500–2000 kcal/week [19
]), an unexpected result given that the amount of activity required to maintain large weight losses is purported to be greater than the activity required to prevent incipient weight gain [18
Our prospective data suggest that an annual change in physical activity equivalent to 1 km/week of running is associated with changes in BMI of −0.015±0.001 and −0.009± 0.001 kg/m2
in men and women, respectively. These estimates are somewhat smaller than the cross-sectional relationships between BMI and km/week of running we have previously reported for men (−0.033±0.001 kg/m2
per km/week) and women (−0.014±0.003 kg/m2
per km/week) [8
]. Others also report that physical activity has a stronger relationship to weight cross-sectionally than to change in weight measured prospectively [20
]. In part, the larger cross-sectional slope may reflect the contributions of self-selection to the cross-sectional relationship. For example, leanness of physically active older women is reported to reflect their leanness during early adulthood (suggesting a component of self-selection) [21
]. In addition, the smaller regression slope of the change data could theoretically be due to greater attenuation of the regression slope by measurement error for change data than cross-sectional data. Specifically, errors in measuring the independent variables are known to bias estimates of the regression slope towards zero. This bias is likely to be greater for change data than cross-sectional data because measurement error is accumulated twice in the calculation of a difference but only once for cross-sectional data. Correcting the regression slope for the apparent measurement error for self-reported running distance would increase the regression slope to −0.024 and −0.015 kg/m2
per Δkm in men and women, respectively, assuming a correlation of 0.89 between repeated measurements [12
Our earlier paper of men studied cross-sectionally suggested that middle-age weight gain is expected if physical activity remains constant, even if the activity is substantial.[8
] We originally estimated that the men would need to increase their distance run by 2.24 km/week annually to compensate for the anticipated weight gain during middle age [8
]. DiPietro et al. have also reported that men and women gained weight during 7.5 years of follow-up unless treadmill test duration improved [22
]. The prospective data presented here suggest that vigorous exercise may need to increase 4.4 km/week annually in men and 6.2 km/week annually in women to compensate for the expected gain in weight due to aging (2.7 and 3.9 km/week annually in men and women, respectively, if we correct for the attenuation due to measurement error associated with self-reported running distance as described above).
The IOM report [6
] concluded that the maintenance of healthy weight (i.e., 18.5 kg/m2
, NHLBI/NIDDK [23
]) requires a level of total energy expenditure that is 160% of basal daily energy expenditure (i.e., a PAL or physical activity index (PAI) of 1.6). Among runners who we estimated to maintain a PAI of 1.7 at both visits, we calculated that the men and women would need to increase their annual weekly running distance by 4.5 and 3.0 km to maintain a constant body weight (analyses not displayed). These estimates are greater than the annual increases of 10 kcal/day in men’s and 7 kcal/day in women’s total energy expenditure that the IOM estimate are required to maintain adult BMIs within the desirable range based on changes in total energy expenditure alone.
We found that changes in weekly running distances had less of an effect on body weight in women than men. Others report that physical activity as measured by doubly labeled water was related to body fat in males but not females [24
]. This finding is unexpected given that the net energy cost of running at self-selected running speeds is reported to be 11% higher in women than men [10
]. Some training studies speculate that the same exercise challenge is less likely to cause weight loss in women than men because women have a greater tendency to compensate for energy expenditure through increased energy intake [26
]. It has also been suggested that training may produce less weight loss in women than men because abdominal fat (generally higher in males) is more responsive to exercise than gluteofemoral fat (generally higher in females) [27
]. BMI is a better predictor of differences in body fat in women than men, so it is unlikely that the difference is due to the inadequacy of BMI to reflect body fat changes in women [6
]. The sex difference may be less apparent for waist circumference than BMI or Δ%body weight because waist circumference is more weakly related to %body fat in women than men [6
The majority of the men and women in our study had BMIs that were below the 25 kg/m2
threshold that the National Institutes of Health and other government and nongovernmental organizations have identified as desirable. However, this does not necessarily mean that increases in BMI below this threshold are benign. Willett et al.[28
] reported that relative to a BMI of 21 kg/m2
, the risk for coronary heart disease (CHD) was 19% higher for women with a BMI of 21–22.9 kg/m2
and 46% higher for a BMI of 23–24.9 kg/m2
. They also reported that weight gain after 18 years of age was a strong predictor of CHD risk even among women whose BMI remained below 25 kg/m2
]. However, others suggest that weight gain does not increase mortality in middle-aged [29
] or older men [31
], or lean postmenopausal women [32
], or that the increased risk primarily restricted to those experiencing the greatest weight gain [33
]. Although the health risks associated with weight gain in the vigorously active men and women remains controversial, their mortality risk is known to be less than sedentary physically unfit individuals matched for weight [34
Our surveys lacked reliable data on changes in energy intake and other sources of energy expenditure that could theoretically account for some of the results reported here. Some of the change in body weight could reflect changes in caloric intake or other activities. Technical limitations of food records and comprehensive activity diaries limit their use in accounting variations in weight over time. Intra-individual variability in daily energy intake is estimated to be ±23% [35
], whereas the long-term error in adjusting cumulative energy intake to expenditure is estimated be less than 2% of energy expenditure [36
]. Underestimation of food intake by food records is reported to range from 10 to 45% [6
]. Between 140 and 700 kcal/day has been attributed to spontaneous physical activities, including fidgeting, which is missed by comprehensive physical activity diaries [37
]. We also note that the duration of follow-up was short (3.2 years in men and 2.6 years in women) and extrapolating our calculations to longer time spans may not be warranted. Our questionnaire requested that the participant report their average running mileage for the current and preceding 4 years, and therefore we lack data on variations in distance, intensity or training schedules that varied in response to season, race preparation, illness or injury.
In our opinion, the more demanding physical activity recommendations by the IOM report represent an important improvement over earlier guidelines.[2
] Our analyses suggest these guidelines may be further improved by: (1) promoting investments in physical activity that increase with age and (2) acknowledging differences in the expected weight loss for men and women who exercise vigorously.