Obesity is a major health concern in the United States and around the world. More than 60% of adult men in the United States and more than 1 billion people worldwide are considered overweight or obese (6,19). Prevalence of obesity is not only high but also increasing. Between 1980 and 2002, the prevalence of obesity among US adults doubled (16). One of the many concerns with this high prevalence of obesity is its association with an increased risk of falls. Fjeldstad et al. (5) reported obese subjects to have a higher prevalence of falls (27% vs 15%) and ambulatory stumbling (32% vs 14%) than their nonobese counterparts. An increased risk of falls increases the risk for injury. Falls were identified as the most common cause of injuries in the obese (∼36% of all injuries) and the cause of a higher proportion of injury-related hospitalization in the obese compared with the nonobese (15).
Weight loss has the potential to improve balance and reduce the risk of falls in the obese. Teasdale et al. (22) investigated the effects of weight loss on balance in obese and morbidly obese subjects. Mean center of pressure (COP) speed during quiet standing decreased linearly with weight loss in both groups. In addition, there were no differences in mean COP speed between obese/morbidly obese subjects after weight loss and nonobese controls. Mean COP speed is considered an index of the activity required to maintain balance (7). Therefore, decreases in weight can result in a decreased amount of activity required to maintain balance and perhaps a decrease in risk of falls. In another study, Maffiuletti et al. (12) investigated the effects of weight loss on balance during single-limb stance. Morbidly obese subjects participated in a body weight reduction program that included an energy-restricted diet, moderate exercise, and nutritional education. Trunk sway, defined as the deviation of the trunk from the medial-lateral axis, decreased with weight loss, and time of balance maintenance, defined as the longest period a subject could stand without help, increased with weight loss. Reduced trunk sway and increased time of balance maintenance may also indicate a decrease in risk of falls.
Strength training also has the potential to reduce risks of falls in the obese (2). Lower extremity strength, specifically at the ankle and knee, can make a significant contribution to balance recovery after a postural perturbation (23). Moreover, the ability to recover balance after a postural perturbation has been shown to improve with increased strength (3,20). Robinovitch et al. (20) investigated the effects of torque development on balance recovery using an ankle strategy. They found that the ability to recover balance after release from a static forward lean is dependent on the capacity to quickly generate and maintain high magnitudes of ankle torque. Although obese adults exhibit increased strength- and power-producing capabilities compared with nonobese, strength and power are significantly lower in obese adults when normalized to body weight (9,10,13). As such, when the obese are perturbed, they may be unable to generate adequate force to recover from the perturbation and are likely to fall. Improving lower limb strength would increase strength relative to body weight, improving balance recovery and perhaps reducing the risk of falls.
Whereas evidence suggests that both weight loss and strength training can improve balance and reduce the risk of falls in the obese, knowing which is more beneficial would allow interventions to be designed to maximize the benefits in balance and reducing risk of falls. Therefore, the goal of this study was to investigate the relative effects of weight loss and strength training on balance recovery using an ankle strategy. Understanding the relative benefit of these interventions on balance recovery (and potentially risk of falls) may improve our understanding of factors that contribute to falls in the obese and contribute to more effective interventions for fall prevention.
Manipulating weight loss and strength independently in human subjects would be difficult and time-consuming. As such, we used a combination of experimental testing and forward dynamic simulations to achieve our goal. Experimental testing involved nine male subjects aged 23.3 ± 4.5 yr (mean ± SD). Subjects had a height of 178.9 ± 6.2 cm, mass of 107.2 ± 10.0 kg, and body mass index (BMI) of 33.5 ± 2.3 kg·m−2. The study was approved by the Virginia Tech Institutional Review Board, and written consent was obtained from all participants before participation.
Experimental testing was performed to determine inputs to the forward dynamic simulations. While harnessed at the waist, subjects were leaned forward with their arms held behind their back and body segments aligned (Fig. 1). Subjects were instructed to recover balance on release using an ankle strategy (defined as contraction of only the muscles spanning the ankle and returning to an upright posture while keeping the body straight and not stepping). This task was chosen because it is a dynamic task associated with fall prevention and it can be modeled relatively easily. The heels were allowed to rise from the ground, and only slight heel raise was used during most trials. The body angle relative to vertical before release, as measured from a line connecting the lateral malleolus and greater trochanter, was defined as the lean angle (θ) (8,17). For the first trial, the lean angle was set to 2° above the subjects' natural lean angle during standing as measured by a goniometer. After each successful recovery, the lean angle was increased by 0.5° and the test repeated. After each failed recovery, another attempt was performed at the same angle. This process was repeated until subjects failed three times at a given angle. A failed recovery was defined as stepping or bending at the hip to recover balance. Similar to previous studies, the maximum lean angle (θmax) was used to quantify balance recovery capability (8,11,17,20).
Subjects were instructed to keep their ankles relaxed while in the forward lean position before release to minimize preparatory plantarflexor muscle activity. In addition, EMG was placed on the lateral gastrocnemius and displayed on an oscilloscope for the researcher to monitor muscle activation. Subjects performed four trials of quiet standing before balance recovery trials, and the average root mean square (RMS) amplitude of the EMG during quiet standing was used as a threshold of muscle activation during the balance recovery trials. If the RMS amplitude rose above this threshold before release, subjects were instructed to relax their ankles.
During all trials, body position was sampled at 100 Hz using a Vicon 460 motion analysis system (Vicon, Lake Forest, CA). Sagittal plane position of the lateral malleolus and greater trochanter were collected and averaged across both sides of the body. Measured from vertical, the line connecting the lateral malleolus and greater trochanter in the sagittal plane was then used to calculate θmax. Ground reaction force was sampled at 1000 Hz using a force platform (Bertec Corporation, Columbus, OH). Voltage applied to the release mechanism that held subjects in the lean position was sampled at 1000 Hz and was used to determine the time of release. Body position, ground reaction force, and release mechanism voltage data were low-pass filtered at 5, 7, and 20 Hz, respectively (eighth-order zero-phase-shift Butterworth filter). In addition, the position of the lateral maleolus and ground reaction force data were used to determine the torque generated about the ankle using an inverse dynamics analysis.
An inverted pendulum model of the human body was used to evaluate the relative effects of weight loss and strength training on balance recovery. This model included a slender rod rotating about a hinge joint to represent the human body rotating about the ankle joints in the midsagittal plane with a torque actuator at the ankle (Fig. 2). Consistent with an earlier study using the same model for the same balance recovery task (20), the center of mass (COM) position (LCOM) was assumed to be located at 50% of the subject's height from the ankles.
A linear approximation of ankle torque history collected during the experimental trials was applied to the model (Fig. 3). Initial ankle torque (Ti), response time (Δt), maximum ankle torque (Tmax), and ankle torque generation rate (C) were calculated using a similar approach as Robinovitch et al. (20). Ankle torque decline rate (D) was defined as the slope of a straight line from Tmax to 500 ms later, and the response time was determined to be the time from release until ankle torque exceeded Ti by 2.5 N·m. Ti, Tmax, C, and D were normalized to the product of body mass (kg) and height (m). Experimental values for θmax and ankle torque properties are given in Figure 3.
The model was validated by correlating the θmax values predicted by the model and θmax values measured experimentally and by determining the RMS error (RMSE) between these same values. To determine the predicted θmax, body mass, height, and ankle torque history parameters of each individual subject were entered into the model. The model was then repeatedly released from progressively increasing angles, similar to our human subject protocol, to determine the maximum lean angle for a successful recovery. Both here and in all subsequent simulations, a successful recovery was defined by the occurrence of θ˙ < 0 while θ < 90° (20).
Multiple simulations (using MATLAB; The MathWorks, Natick, MA) were used to compare the potency of three interventions for increasing θmax. The first intervention, termed "Tmax," involved increasing the maximum torque producing capability of the ankle plantarflexors. Strength training the ankle plantarflexors increases not only their maximum torque producing capability but also their torque generation rate (4). As such, the second intervention, termed "strength training," involved increasing Tmax and the torque generation rate C based on the linear relation between increases in maximum torque and torque generation rate as reported by Del Balso and Cafarelli (4). The third intervention, termed "weight loss," involved decreasing body mass and varying body COM position (LCOM) and radius of gyration (rCOM) according to Matrangola et al. (14). Although this previous study only reported changes in body segment inertial parameters with approximately 14% weight loss, the current study assumed LCOM and rCOM changed linearly with weight loss and were interpolated for varying amounts of weight loss. Body mass and Tmax were manipulated in intervals of 0.01% (approximately 0.01 kg) and 0.05% (approximately 0.1 N·m) from initial, respectively.
A two-way repeated-measures ANOVA was used to compare the percent changes in weight, Tmax, and strength (includes changes in Tmax and C) needed to achieve prescribed improvements in θmax. As such, our independent variables were as follows: intervention (Tmax, strength training, and weight loss) and improvement in θmax (0°, 0.5°, 1.0°, 1.5°, and 2.0°). The dependent variable was percent change in Tmax, strength, or weight loss necessary to achieve these prescribed improvements in θmax. Pairwise comparisons were evaluated using Tukey's Honestly Significant Difference (HSD). A log10 transformation was applied to the data to obtain a normal distribution for the statistical analysis, but nontransformed data were presented in the results for clarity. Statistical analyses were performed using JMP v7 (Cary, NC), and a significance level of P < 0.05 was used.
The balance recovery model predicted the experimental θmax values with reasonable accuracy (Fig. 4). The coefficient of determination between predicted and measured θmax values was r2 = 0.83, and the RMSE was 0.76°.
A contour plot was created (Fig. 5) to illustrate the relative effects of strength training and weight loss on balance recovery for an "average" subject (all model parameters were set to mean experimental values). Although both strength training and weight loss increased θmax, the percent changes in strength that was required to increase θmax by 1° were nearly double that of weight loss. If strength training and weight loss were applied simultaneously, smaller changes in each can be used to achieve the same increase in θmax.
Using results from all subjects (Fig. 6), significant effects were found for intervention (P < 0.001), θmax (P < 0.001), and their interaction (P = 0.002). Pairwise comparisons indicated statistically significant differences among all three interventions at all levels of θmax investigated. Weight loss consistently required the smallest change of all three interventions to achieve a desired increase in θmax, and increases in Tmax alone consistently required the largest change. For example, to achieve a 1° increase in θmax, an 8.6 ± 0.8% decrease in weight, a 15.3 ± 1.1% increase in strength, or a 22.8 ± 5.7% increase in Tmax were required. The differences in efficacy among the three interventions became more apparent as larger increases in θmax were targeted.
The purpose of this study was to evaluate the relative effects of weight loss and strength training on balance recovery using an ankle strategy. Although all three interventions improved balance recovery, a targeted improvement in balance recovery could be achieved with a smaller amount of weight loss than increase in strength. For example, a 1% reduction in weight had an equivalent improvement on θmax as a ∼1.8% increase in strength over the range of θmax investigated in this study. Alternatively, a 1-kg reduction in body mass had an equivalent improvement on θmax as a ∼3.2-N·m improvement in ankle strength (including a concomitant increase in torque generation rate). An increase to Tmax alone required significantly higher increases to improve θmax. In addition to these findings, the differences between these interventions became more apparent as larger improvements in balance were targeted. These data suggest that weight loss is a more potent intervention than strength training in improving balance recovery using an ankle strategy.
The average experimental θmax in this study (9.8 ± 1.3°) was higher than that reported by Robinovitch et al. (20) (5.9 ± 1.1°), for a group of young adults aged 17-38 yr, and by Mackey and Robinovitch (11) (7.2 ± 1.2°), for a group of young women aged 19-36 yr, all performing a similar task. The higher experimental θmax in the present study may have resulted from at least two differences between these studies. First, the present study measured θ as the angle between vertical and a line connecting the ankle and greater trochanter, whereas both previous studies measured θ as the angle between vertical and a line connecting the ankle and shoulder. This may account for slight systematic differences in θmax. Second, the present study included only young males aged 19-32 yr (with only two subjects older than 23 yr) who may exhibit higher θmax values, on average, compared with the slightly older and/or female subjects included in the previous studies (11).
The balance recovery model predicted θmax with reasonable accuracy but tended to underestimate θmax (Fig. 4). There were several differences between our model and our human subjects that could have contributed to this error. First, the linear approximation of ankle torque used in the model typically underestimated experimental ankle torque generated during recovery (average error = −1.53 N·m·s) as determined by integrating the area between these two timehistories. Applying a slightly smaller angular impulse to the model than in our experiments would lead to a smaller total change in angular momentum because of impulse-momentum relation and negatively influence balance recovery limits. Second, small amounts of hip movement during recovery would contribute to shear forces at the floor (i.e., hip strategy) and likely facilitate recovery in our subjects compared with a purely ankle strategy used by the model. Third, our inverted pendulum model consisted of a slender rod, similar to a study conducted by Robinovitch et al. (20), which implicitly assumes that the subject's mass was uniformly distributed along the subject's height. This model was chosen for simplicity and because of a lack of guidance on COM position for the obese in the literature. Adipose tissue distribution is highly variable (1) and can influence inertial parameters in the obese. However, we think that our results pertaining to the relative benefits of the interventions investigated were largely insensitive to COM location. Fourth, our inverted pendulum model did not include heel rise, and this could account for some differences between our predicted and experimental values of θmax. Despite these differences between our model and experimental data and subjects, we think that it is reasonable to generalize our modeling results to human subjects for the specific task investigated.
Two limitations to this study warrant discussion. First, the balance recovery task involved releasing subjects from a static forward lean, whereas most falls do not occur from a stationary body position (18). It is unclear how the results from this study would apply to these dynamic balance recovery tasks. More importantly, θmax was used as a surrogate measure of balance recovery capability and the relation between θmax and risk of falls has not been determined. It would seem, however, that the maximum increase in θmax of 2° reported here is clinically significant because a previous study (11) reported a 2.6° difference in θmax between young women (mean age = 25 yr) and older women (mean age = 78 yr). A second limitation is that it is unclear how well our results generalize to individuals outside the BMI range examined in this study.
This study quantified changes in weight and strength needed to improve balance recovery capability, but these results do not give insight on which intervention may be easier for individuals to achieve. For example, the contour plot (Fig. 5) indicates that a ∼9% decrease in weight or a ∼15% increase in strength would improve θmax by 1°. However, it may be easier for some individuals to achieve a 15% increase in strength rather than a 9% decrease in weight. It is likely that the difficulty of achieving these changes would be highly dependent on the specific interventions used and differences between individuals. Results from literature, however, can be used to begin to determine which intervention goal may be more achievable. Sartorio et al. (21) completed a study investigating the effects of a weight loss intervention on muscle strength and power. The subjects participated in a 3-wk program that combined exercise (aerobic and strength training) and energy-restricted diet. BMI of the male subjects decreased from 41.3 ± 4.0 to 39.1 ± 3.7 kg·m−2, corresponding to 5.1 ± 0.9% weight loss, and lower limb strength, as measured by a leg press exercise, increased by 32.3 ± 24.5%. Applying these changes to our "average" subject model (Fig. 5), θmax would increase by 0.6° because of weight loss and by 2.2° because of strength training. This indicates that the improvement in θmax from strength training was larger than that from weight loss, although θmax is more sensitive to changes in weight, simply because larger changes in strength were achieved. This suggests that strength gain may actually be a more beneficial intervention for reducing risk of falls in the obese in practice, but further work is needed to address this more conclusively.
In conclusion, a targeted improvement in balance recovery can be achieved with a smaller amount of weight loss than increase in strength. This suggests that weight loss is a more potent intervention than strength training in improving balance recovery using an ankle strategy. However, the relative benefits of these interventions are less clear when considering the human factors that influence the achievability of weight loss and strength gain. Further experimental research is needed before a more definitive recommendation can be ascertained to help prevent falls in the obese.
Disclosure of funding: SLM was supported by a National Defence Science and Engineering Graduate (NDSEG) Fellowship.
Conflict of interest: The authors report no conflict of interest.
The results of the present study do not constitute endorsement by ACSM.
1. Björntorp P. Adipose tissue distribution and function. Int J Obes
. 1991;15(Suppl 2):67-81.
2. Clark K. Balance and strength training for obese individuals. ACSM's Health Fit J
3. Corbeil P, Simoneau M, Rancourt D, Tremblay A, Teasdale N. Increased risk for falling associated with obesity: mathematical modeling of postural control. IEEE Trans Neural Syst Rehabil Eng
4. Del Balso C, Cafarelli E. Adaptations in the activation of human skeletal muscle induced by short-term isometric resistance training. J Appl Physiol
5. Fjeldstad C, Fjeldstad AS, Acree LS, Nickel KJ, Gardner AW. The influence of obesity on falls and quality of life. Dyn Med
6. Flegal KM, Carroll MD, Ogden CL, Johnson CL. Prevalence and trends in obesity among US adults, 1999-2000. JAMA
7. Geurts AC, Nienhuis B, Mulder TW. Intrasubject variability of selected force-platform parameters in the quantification of postural control. Arch Phys Med Rehabil
8. Grabiner MD, Owings TM, Pavol MJ. Lower extremity strength plays only a small role in determining the maximum recoverable lean angle in older adults. J Gerontol A Biol Sci Med Sci
9. Hulens M, Vansant G, Lysens R, Claessens AL, Muls E, Brumagne S. Study of differences in peripheral muscle strength of lean versus obese women: an allometric approach. Int J Obes Relat Metab Disord
10. Lafortuna CL, Maffiuletti NA, Agosti F, Sartorio A. Gender variations of body composition, muscle strength and power output in morbid obesity. Int J Obes (Lond)
11. Mackey DC, Robinovitch SN. Mechanisms underlying age-related differences in ability to recover balance with the ankle strategy. Gait Posture
12. Maffiuletti NA, Agosti F, Proietti M, et al. Postural instability of extremely obese individuals improves after a body weight reduction program entailing specific balance training. J Endocrinol Invest
13. Maffiuletti NA, Jubeau M, Munzinger U, et al. Differences in quadriceps muscle strength and fatigue between lean and obese subjects. Eur J Appl Physiol
14. Matrangola SL, Madigan ML, Nussbaum MA, Ross R, Davy KP. Changes in body segment inertial parameters of obese individuals with weight loss. J Biomech
15. Matter KC, Sinclair SA, Hostetler SG, Xiang H. A comparison of the characteristics of injuries between obese and non-obese inpatients. Obesity (Silver Spring)
16. Ogden CL, Carroll MD, Curtin LR, McDowell MA, Tabak CJ, Flegal KM. Prevalence of overweight and obesity in the United States, 1999-2004. JAMA
17. Owings TM, Pavol MJ, Foley KT, Grabiner MD. Measures of postural stability are not predictors of recovery from large postural disturbances in healthy older adults. J Am Geriatr Soc
18. Pai Y-C, Patton J. Center of mass velocity-position predictions for balance control. J Biomech
19. Puska P, Nishida C, Porter D. Global Strategy on Diet, Physical Activity and Health. Obesity and Overweight [Internet]. World Health Organization (WHO); [cited 2008 Sept 10]. Available from: http://www.who.int
20. Robinovitch SN, Heller B, Lui A, Cortez J. Effect of strength and speed of torque development on balance recovery with the ankle strategy. J Neurophysiol
21. Sartorio A, Maffiuletti NA, Agosti F, Lafortuna CL. Gender-related changes in body composition, muscle strength and power output after a short-term multidisciplinary weight loss intervention in morbid obesity. J Endocrinol Invest
22. Teasdale N, Hue O, Marcotte J, et al. Reducing weight increases postural stability in obese and morbid obese men. Int J Obes (Lond)
23. Whipple RH, Wolfson LI, Amerman PM. The relationship of knee and ankle weakness to falls in nursing home residents: an isokinetic study. J Am Geriatr Soc