Effects of Exercise Involving Predominantly Either Joint-Reaction or Ground-Reaction Forces on Bone Mineral Density in Older Women

Authors

  • Wendy M. Kohrt,

    Corresponding author
    1. Washington University School of Medicine, Division of Geriatrics and Gerontology, Department of Internal Medicine, St. Louis, Missouri, U.S.A.
    • Wendy M. Kohrt, Ph.D. Washington University School of Medicine 660 S. Euclid, Box 8113 St. Louis, MO 63110 U.S.A.
    Search for more papers by this author
  • Ali A. Ehsani,

    1. Washington University School of Medicine, Division of Geriatrics and Gerontology, Department of Internal Medicine, St. Louis, Missouri, U.S.A.
    Search for more papers by this author
  • Stanley J. Birge JR.

    1. Washington University School of Medicine, Division of Geriatrics and Gerontology, Department of Internal Medicine, St. Louis, Missouri, U.S.A.
    Search for more papers by this author

Abstract

This study compared the effects of two exercise training programs, 11 months in duration, on bone mineral density (BMD) in older, sedentary women. Thirty-nine women, aged 60–74 years, were assigned to the following groups: (a) a group that performed exercises that introduced stress to the skeleton through ground-reaction forces (GRF) (i.e., walking, jogging, stairs); (b) a group that performed exercises that introduced stress to the skeleton through joint-reaction forces (JRF) (i.e., weight lifting, rowing); or (c) a no-exercise control group. BMD of the whole body, lumbar spine, proximal femur, and distal forearm was assessed five times at ∼3-month intervals. The GRF and JRF exercise programs resulted in significant and similar increases in BMD of the whole body (2.0 ± 0.8% and 1.6 ± 0.4%, respectively), lumbar spine (1.8 ± 0.7% and 1.5 ± 0.5%, respectively), and Ward's triangle region of the proximal femur (6.1 ± 1.5% and 5.1 ± 2.1%, respectively). There was a significant increase in BMD of the femoral neck only in response to the GRF exercise program (GRF, 3.5 ± 0.8%; JRF, −0.2 ± 0.7%). There were no significant changes in BMD in control subjects. Among all exercisers, there was a significant inverse (r = −0.52, p < 0.01) relationship between increases in whole body BMD and reductions in fat mass, suggesting a dose response effect of exercise on bone mass. Although femoral neck BMD was responsive only to the GRF exercise program, some adaptations (i.e., increase in lean body mass and strength) that were specific to the JRF exercise program may be important in preventing osteoporotic fractures by reducing the risk for falls. It remains to be determined whether all of these benefits can be gained through a training program that combines the different types of exercises employed in this study.

INTRODUCTION

Although it has become axiomatic that habitual exercise helps prevent osteoporosis, little is known regarding the type of exercise most likely to be beneficial. Studies in animals indicate that the osteogenic response to dynamic compressive force is positively related both to the magnitude of the force applied1 and to the rate of loading2 and that very few loading cycles are necessary to generate a maximal response.3 Furthermore, the bone modeling response appears to depend on whether the loading forces introduce an atypical stress to the bone, that is, one that is novel in its magnitude and/or distribution of strain.1 These findings suggest that exercises that produce relatively high strains that are not performed as part of daily activities, such as heavy resistance training, may be more effective in inducing bone mineral accretion than repetitious routine activities, such as walking.

Prospective studies indicate that resistance exercise training can induce small, but significant, increases in bone mineral density (BMD).4–6 However, the increases do not appear to be greater than those that can be induced by relatively high intensity endurance-type exercise training.6,7 There are few direct comparisons of the effectiveness of different types of exercise in bringing about increases in bone mass, particularly in older women at risk for osteoporosis. We therefore compared the effects of two different exercise training programs, distinguished by the manner in which forces acting on the skeleton were generated, on BMD in older, sedentary women. One training program involved exercises that introduced stress to the skeleton through ground-reaction forces (GRF) (i.e., walking, jogging, stair climbing), while the other program included activities that introduced stress to the skeleton through joint-reaction forces (JRF) (i.e., weight lifting, rowing).

MATERIALS AND METHODS

Subjects

The subjects for this study were 48 healthy women, 60–74 years of age, who did not engage in regular exercise. Assignment to the three study groups (two exercise groups, one control group) was done by matching for body weight. Data are reported for the 39 women who completed follow-up testing. Of the nine women who did not complete the study, two were in the GRF exercise program (one because of the time commitment; one because of chronic knee pain), four were in the JRF exercise program (one moved out of state; one was injured in a fall on ice; one required oral surgery and decided not to continue; one because of chronic back pain), and three were control subjects (one started an exercise program; one went on a diet and lost 9 kg; one was diagnosed with lung cancer).

Eleven of the participants had used estrogen previously for an average duration of 3.4 ± 1.1 years (mean ± SE), but none had been taking estrogen for at least 2 years. All of the participants were nonsmokers. Subjects provided written informed consent to participate in the study, which was approved by the Human Studies Committee of the Washington University School of Medicine.

Preliminary screening tests included a medical history, physical examination, chest X-ray, graded treadmill exercise test with blood pressure and electrocardiogram monitoring, and blood and urine chemistries, including measures of serum calcium concentration and 24-h urine calcium and creatinine excretion. Chemistries were performed using standard automated laboratory techniques. Subjects were excluded from the study if they had any medical problems that contraindicated exercise or prevented performance of vigorous exercise.

Diet evaluation and calcium supplementation

Subjects completed 7-day food records during the initial screening and again at the completion of the study. Foods were weighed and recorded in household measures. A registered dietician instructed the subjects on the procedures for recording food intake and also conducted interviews shortly after the records were obtained to validate the accuracy of the portions that were recorded. The food intake records were analyzed using Nutritionist IV (N-Squared Computing, Salem, OR, U.S.A.). Because adequate calcium intake has been shown to be important for realizing the beneficial effects of exercise on BMD,8 calcium intake was adjusted to ∼1500 mg/day in all participants after the initial dietary assessment. Calcium intake was monitored by having the women complete every 3 months a 7-day record of their intake of dairy products, other major calcium-containing foods (from a list provided), calcium supplements, antacids, and multivitamins.

Exercise programs

Subjects in the two exercise treatment groups were required to attend three exercise sessions per week and were encouraged to attend five sessions per week. The first 2 months of training were the same for both treatment groups. This phase of the training program focused on exercises designed to improve range of motion and flexibility of all major joints and muscle groups, with the intent of reducing the likelihood of injury in the subsequent, more vigorous 9-month exercise programs.

The two 9-month exercise programs were distinguished by the means by which stress was introduced to the skeleton. The GRF exercise program included activities in which forces acting on the skeleton were generated by ground-reaction forces (i.e., walking, jogging, stair climbing), whereas the JRF exercise program included activities that involved predominantly joint-reaction forces (i.e., weight lifting, rowing). In both programs, exercise prescriptions were individualized and updated on a weekly basis.

The initial goal for participants in the GRF exercise program was to walk for 30 minutes (not including warm up and cool down) at a moderate intensity (i.e., 60–70% of maximal heart rate). Thereafter, the rates at which duration and intensity of exercise were increased depended on the extent to which an individual could tolerate increases in aerobic and orthopedic demands. The long-term goal was to exercise at least 45 minutes/day at an average intensity of 80–85% of maximal heart rate (HR). At the beginning of the third month, stair climbing/descending and/or jogging were incorporated into the training protocol. All participants were strongly encouraged to jog at least intermittently (e.g., jog one out of every three laps) to increase the magnitude of the ground reaction forces,9 thereby increasing the peak forces acting on the lumbar spine10 and proximal femur.11

In the JRF exercise program, approximately one-half of each exercise session was devoted to weight lifting and one-half to rowing. Weight lifting exercises utilized free weights (overhead press, biceps curl, triceps extension), Hoist and Nautilus equipment (leg press, leg extension, leg flexion, bench press), and a Smith machine with a counterbalanced bar (squats). In some cases, exercises could not be performed (e.g., squats) or had to be modified (e.g., incline press rather than overhead press) because of orthopedic discomfort. Exercises using free weights were performed in a standing posture whenever possible to increase the loading forces on the spine and hip. During the first week, participants became familiar with the equipment. The maximal weight that could be lifted one time (1-RM) was then determined for all exercises except the squat. Subjects were prescribed two to three sets of each exercise two times per week at an intensity that resulted in fatigue after 8–12 repetitions. Resistance was increased whenever a participant was able to complete the prescribed number of sets and repetitions. The 1-RMs were re-evaluated at ∼3-month intervals. The initial goal of the rowing component of the exercise program was to exercise for 15–20 minutes at a moderate intensity (60–70% of maximal HR). Thereafter, intensity was gradually increased so that participants performed two or three 10-minute bouts at an intensity that was 80–85% of maximal HR.

To determine if there was a dose-response relationship between exercise and bone mineral accretion, the amount of exercise performed was compiled on a weekly basis for each individual. This was done by summing the total amount of weight lifted on each exercise for the weight-lifting portion of the JRF exercise program and by estimating the caloric expenditure for the rowing portion of the JRF exercise program and for all activities in the GRF exercise program. The equations recommended by the American College of Sports Medicine were used to estimate energy expenditure during walking, jogging, and stair climbing.12 Energy expenditure during rowing was estimated using the equation recommended by the manufacturer of the rowing ergometers (Concept II, Morrisville, VT, U.S.A.).

Maximal aerobic power

Maximal aerobic power (V̊O2max) was assessed during graded treadmill walking as described previously.13 Subjects met at least two of the following criteria for the attainment of V̊O2max: plateau in V̊O2, HR within 10 beats of the age-predicted maximal HR, and respiratory exchange ratio greater than 1.10. V̊O2max was assessed before and after the flexibility exercise program, and at 3-month intervals through the exercise programs. In nonexercisers, V̊O2max was assessed at the onset and completion of the study.

Muscle strength

Peak torque of the quadriceps and hamstrings was evaluated in all participants on a Cybex II dynamometer at the beginning and end of the study. Participants were seated with the back supported and the hips at 120° of flexion with the dominant leg strapped to the movement arm of the dynamometer. The movement arm was individually adjusted for leg length and the knee joint axis of motion. Tests were performed at 60 and 120°/s for each muscle group. Three trials were performed at each speed, and the best two trials were averaged to obtain peak torque. A composite strength score, which is the average percent increase for both muscle groups at both test speeds, is reported. For participants in the joint-reaction force exercise program, increases in strength were also evaluated by assessment of the 1-RMs.

BMD and body composition

BMD of the proximal femur (neck, trochanter, and Ward's triangle), lumbar spine (L2–L4), wrist (ultradistal and one-third distal regions), and total body was measured by dual X-ray absorptiometry (DXA) using the Hologic QDR-1000/W instrument (Hologic, Waltham, MA, U.S.A.). In 60− to 70-year-old women, the CV for BMD was ≤1.8% at all sites of interest except the Ward's triangle region of the proximal femur, which had a CV of 3.2%.14

BMD was assessed before and after the flexibility exercise program and at approximately 3-month intervals through the exercise programs, for a total of five assessments. Subjects in the control group also had BMD assessed at 3-month intervals for 1 year. To minimize variability in BMD due to technical sources (e.g., positioning) and to standardize the time interval over which measures were obtained, a regression equation was generated for each individual and each measurement site to determine the change in BMD per year (g·cm−2·year−1). The total body DXA was used to assess changes in body composition over the study, using version 5.64 of the enhanced whole body software.

Bone turnover rate was evaluated by measuring serum osteocalcin levels at baseline and at the end of the protocol. Because of diurnal variations in osteocalcin, all blood samples were drawn in the morning after an overnight fast, processed immediately, and stored at −80°C for subsequent batch analysis by radioimmunoassay.15

Statistical analyses

One-way analysis of variance (ANOVA) and the Student-Newman-Keuls pairwise comparison post hoc test were used to detect differences among the groups in baseline measures. One-way ANOVAs were performed to determine if there were differences among the groups in the changes that occurred during the treatment/observation period. The changes in BMD within each group were also evaluated by paired t-tests to determine if they were significantly different from zero. Zero-order correlations between changes in BMD and determinants of the amount of exercise performed (e.g., exercise frequency, duration, and intensity; estimated caloric expenditure; amount of weight lifted) or the physiological adaptations to training (e.g., increase in strength and maximal aerobic power; changes in body composition) were evaluated to determine whether there was a dose-response effect of exercise on bone. Statistical significance was defined as an alpha level at or below 0.05, and all data are reported as mean ± SE unless otherwise stated.

RESULTS

Exercise training

Attendance averaged 3.4 ± 0.2 exercise sessions per week in both of the exercise groups. During the first 4 weeks of the GRF exercise program, subjects exercised an average of 33 ± 2 minutes/day at an intensity of 73 ± 2% of maximal HR; the estimated energy expenditure was 690 ± 73 kcal/week. By the last 4 weeks of the program, the women in this group were exercising an average of 52 ± 3 minutes/day at an intensity of 81 ± 2% of maximal HR; the estimated energy expenditure was 1253 ± 160 kcal/week.

For the rowing portion of the JRF exercise program, participants exercised an average of 24 ± 2 minutes/day at an intensity of 74 ± 1% of maximal heart rate during the first 4 weeks of the study, progressing to 30 ± 4 minutes/day at an intensity of 83 ± 1% of maximal heart rate during the final 4 weeks. The estimated energy expenditure increased from 589 ± 53 kcal/week during the first month to 871 ± 143 kcal/week during the final month of the study. For the weight lifting portion of the program, the average number of sets performed per week, number of repetitions per set, and amount of weight lifted per repetition for each lifting exercise during the third and ninth months of the study are presented in Table 1.

Table Table 1. Average (Mean ± SE) Number of Sets Performed per Week, Repetitions per Set, and Weight Lifted per Repetition During Training Sessions in the Third and Ninth Months of the Joint-Reaction Force Exercise Program
original image

Body composition

At the onset of the study, the groups were similar in age, age of onset of menopause, height, weight, fat-free mass (FFM), and body fat content (Table 2). There was a significant reduction in body weight (p < 0.01) in response to the GRF exercise program only, but body fat content was similarly reduced (p < 0.01) in response to both exercise programs (Table 3). Only the JRF exercise program resulted in a significant increase in FFM (p < 0.01). Control subjects had no significant changes in body composition over the period of study.

Table Table 2. Baseline Characteristics of Study Groups
original image
Table Table 3. Changes in Body Composition, Maximal Aerobic Power, and Serum Chemistries in Treatment and Control Groups
original image

Muscle strength

Muscle strength of the leg extensors and flexors, assessed on an isokinetic dynamometer, increased by 15 ± 5% (p < 0.01) in response to the JRF exercise program and by 9 ± 4% (p < 0.05) in response to the GRF exercise program. The change in muscle strength in the control group was not significant (7 ± 6%). In the JRF exercise group, the increases in the 1-RMs in response to training were all highly significant (all p < 0.01): leg press, 35 ± 10% (from 95 ± 8 to 122 ± 7 lb); leg extension, 54 ± 11% (from 76 ± 7 to 118 ± 14 lb); leg flexion, 54 ± 10% (from 57 ± 8 to 89 ± 15 lb); bench press, 78 ± 14% (from 39 ± 6 to 66 ± 6 lb); shoulder press, 70 ± 10% (from 25 ± 3 to 41 ± 5 lb); biceps curl, 57 ± 11% (from 22 ± 3 to 32 ± 4 lb); triceps press, 48 ± 2% (from 9 ± 1 to 13 ± 1 lb).

Maximal aerobic power

Initial V̊O2max values were similar among the groups (Table 2). By the end of the study, V̊O2max was significantly increased in both of the exercise groups (GRF exercise, 3.6 ± 0.9 ml/minute/kg; JRF exercise, 2.5 ± 0.7 ml/minute/kg; both p < 0.001), but not in the control group (−1.2 ± 0.5 ml/minute/kg).

Diet

Total energy intake and the macronutrient composition of the diet were similar among the groups at the beginning of the study. Energy intake averaged 1800 ± 51 kcal/day, and the diet was composed of 16 ± 1% protein, 51 ± 1% carbohydrate, 30 ± 1% fat, and 3 ± 1% alcohol. There were no significant changes in these measures in any of the groups during the study. Calcium intake was not different among the groups prior to supplementation and increased similarly in all groups with supplementation. Average calcium intake at the beginning and end of the study was 1050 ± 82 mg/day and 1440 ± 60 mg/day, respectively (p < 0.01).

Blood and urine chemistries

Initial fasted serum and urine calcium concentrations were similar among the groups, averaging 9.5 ± 0.1 mg/dl and 10.0 ± 1.3 mg/dl, respectively, in all subjects. At the end of the study, the serum calcium concentration was reduced to a similar extent in all three groups (to 9.3 ± 0.1 mg/dl, p < 0.01), but the urine calcium concentration was unchanged (9.9 ± 0.8 mg/dl). The 24-h urine calcium-to-creatinine ratio at the beginning of the study was lower in women in the control group (0.11 ± 0.01) than in those in the JRF (0.20 ± 0.01, p < 0.01) and GRF (0.17 ± 0.03, p < 0.05) exercise groups. The ratio tended to increase in response to calcium supplementation in control subjects (to 0.15 ± 0.02, p = 0.06), but was unchanged in the two exercise groups (JRF, 0.21 ± 0.02; GRF, 0.18 ± 0.02).

Bone mineral density

Prior to intervention, there were no significant differences among the groups in BMD at any of the sites measured (Table 4). Both exercise programs resulted in significant increases in BMD of the whole body (GRF, 2.0 ± 0.8% JRF, 1.6 ± 0.4%), lumbar spine (GRF, 1.8 ± 0.7%; JRF, 1.5 ± 0.5%), and Ward's triangle region of the proximal femur (GRF, 6.1 ± 1.5%; JRF, 5.1 ± 2.1%) (Fig. 1). In addition, there was a significant increase in BMD of the femoral neck (3.5 ± 0.8%) in response to the GRF exercise program, but not the JRF program (−0.2 ± 0.7%). The magnitude of increase in BMD of Ward's triangle was significantly (p < 0.01) and inversely (r = −0.69) related to the initial BMD ratio of the Ward's triangle region relative to the femoral neck region. In other words, the largest increases in Ward's triangle BMD occurred in those subjects in whom this region was low relative to the femoral neck BMD. Neither of the exercise programs had a significant effect on BMD of distal forearm regions. There were no significant changes in BMD at any of the sites measured in control subjects over the period of study.

Table Table 4. Baseline Bone Mineral Density
original image
Figure FIG. 1.

Absolute (top panel) and relative (bottom panel) changes in bone mineral density (BMD) in response to ground-reaction force and joint-reaction force exercise training programs and in no-exercise control subjects. Different from control group, *p < 0.05, †p < 0.01.

There were no significant relationships between the estimates of the amount of exercise performed during the supervised exercise programs and the magnitude of change in BMD for any of the skeletal sites measured. There were, however, significant inverse relationships between the increase in whole body BMD and the reductions in body weight (r = −0.37, p < 0.05) and fat mass (Fig. 2; r = −0.52; p < 0.01) that occurred in response to exercise. Since there was no change in energy intake in exercising subjects (initial, 1825 ± 65 kcal/day; final, 1818 ± 70 kcal/day), the decrease in fat mass reflected the increase in physical activity level during the period of study.

Figure FIG. 2.

Changes in whole body bone mineral density (BMD) relative to changes in fat mass that occurred in response to exercise training.

Bone turnover

The initial serum osteocalcin concentration was higher in subjects in the JRF exercise group (29.6 ± 3.7 ng/ml) than in subjects in the control (21.5 ± 3.2 ng/ml, p < 0.05) and GRF exercise groups (22.4 ± 2.3 ng/ml, p = 0.08). The osteocalcin level was reduced in response to the JRF exercise program (to 21.8 ± 2.2 ng/ml, p < 0.05) to a level that was not different from final values in the other two groups (GRF exercise, 22.3 ± 2.1 ng/ml; control, 19.8 ± 2.2 ng/ml). There were no significant relationships between osteocalcin levels and BMD, either before or after exercise training, and the changes in osteocalcin in response to exercise were not predictive of changes in BMD.

DISCUSSION

The principle finding of this study was that two different exercise training programs, which introduced stress to the skeleton through either ground-reaction or joint-reaction forces, both resulted in significant increases in BMD at clinically important sites in older postmenopausal women. There were significant and similar increases in BMD of the whole body, lumbar spine, and Ward's triangle region of the proximal femur in response to both training programs. However, at the femoral neck, which is the fracture site associated with a high degree of morbidity and mortality, only the GRF exercise program brought about an increase in BMD. The JRF exercise program was sufficient only in preventing mineral loss at the femoral neck.

Because Ward's triangle is the computer-generated region of lowest density within the femoral neck region, it was somewhat surprising that a significant increase in BMD of Ward's triangle occurred in response to the JRF exercise program in the absence of an increase at the femoral neck. However, this is not a unique observation. In a study by Kerr et al.,5 high intensity resistance exercise training resulted in a 2.3% increase in BMD of Ward's triangle, but no increase in femoral neck BMD, in postmenopausal women. Similarly, in young women, Lohman et al.4 found that 1 year of resistance exercise training resulted in a 1.8% increase in Ward's triangle BMD despite a slight decrease in BMD of the femoral neck.4 These observations suggest that exercise-induced modeling/remodeling of bone within the femoral neck area specifically targets the weakest regions. In the current study, the largest increases in BMD of Ward's triangle in response to exercise occurred in those women whose baseline values were low relative to the BMD of the femoral neck region. The notion that exercise targets the weakest regions of the skeleton seems logical, given the contention that mechanical loading is the only stimulus capable of producing mechanically and structurally appropriate increases in bone mass.16

Based on the work of Rubin and colleagues, it is likely that the magnitude of the peak loading forces acting on specific regions of the skeleton is a major determinant of whether the region will undergo modeling/remodeling.1,3 Peak ground-reaction forces during walking and running are approximately 1–1.5 times and 2–3 times body weight, respectively.9 Telemetry data from patients fitted with instrumented hip prostheses indicate that peak joint forces during walking are 2.8–4.8 times body weight at walking speeds ranging from 1–5 km/h.11,17–19 Fast walking, jogging, and going up and down stairs all appear to generate peak hip joint forces of similar magnitude, with an upper limit of 5–6 times body weight.11,17–19 Thus, it is likely that the peak hip joint forces generated during the GRF exercise program were 5–6 times body weight. Because comparable data are apparently not available for weight lifting and rowing activities, it is difficult to estimate the magnitude of hip joint forces generated by these activities. Compressive forces on the knee joint during the squat exercise have been reported to be ∼6 times body weight, suggesting that relatively high forces may also occur at the hip.20 However, the squat exercise is unique in that it utilizes body weight in addition to the external load being lifted. It is unlikely that the other lower extremity weight-lifting exercises, which involved only an external load, generated similar magnitudes of force. The potential of weight-lifting exercises to have an impact on bone modeling/remodeling of the hip probably depends on a number of factors, including the magnitude of the load, the position of the lower extremities and the trunk, the position of the load relative to the extremities, and the speed of movement.20 Moreover, the complexity of these factors probably contributes to the lack of uniformity in the literature regarding the effects of exercise on BMD of the hip. In contrast to our findings, exercises that introduce stress to the skeleton through ground-reaction forces do not necessarily result in increases in femoral neck BMD,21,22 while weight lifting has been found by others to result in significant increases in femoral neck BMD.23,24

The classic in vivo loading studies of Lanyon, Rubin, and colleagues have demonstrated that the osteogenic response is maximized by just a few loading cycles3 when the loading forces are high,1 applied at a fast rate,2 and result in a nonconventional strain distribution.1 It is important to point out that the two exercise programs employed in this study differed in several respects with regard to these factors. The GRF exercises probably involved higher peak and average loading forces and faster loading rates than the JRF exercises. Also, because the GRF exercises involved ambulation, every loading force was targeted to weight-bearing regions of the skeleton. However, since these activities (i.e., walking and stairs) were probably performed by the participants on a daily basis, the loading forces were atypical only in terms of magnitude, not distribution of strain. On the other hand, it is likely that the forces produced during the JRF exercises, although lower in magnitude and loading rate, represented a more unique stress to the skeleton, since the activities (i.e., weight lifting and rowing) were novel to all of the participants. Our finding that novel exercises involving relatively lower strains and strain rates can induce similar osteogenic responses as more conventional exercises involving higher strains and strain rates supports Lanyon's minimum effective strain-related stimulus theory of the adaptive response of bone to functional loading.16

Extensive records of the amount of exercise performed were maintained for each participant with the intent of determining whether there was a dose-response effect between the amount of exercise performed and the change in BMD. None of the parameters, which included exercise frequency, duration, and intensity, estimated total caloric expenditure, and total amount of weight lifted, were predictive of changes in BMD at any of the sites measured. Among all exercisers, however, there were significant inverse relationships between the increase in whole body BMD and the reductions in body mass and fat mass that occurred in response to exercise. We believe that this is indicative of a dose-response relationship between exercise and bone modeling/remodeling for two reasons. First, since food intake did not change significantly over the period of study, the decrease in fat mass is an estimate of the net increase in energy expenditure (i.e., physical activity) during the period of study. Second, the finding that the largest increases in BMD occurred in women who had the greatest reductions in body weight and fat mass runs counter to the expected change, because reductions in bone mass typically accompany weight loss.25–28 Jensen and colleagues25 found that there was a loss of 16.5 g of bone mineral for each 1 kg reduction in fat mass and further demonstrated that this was not due to technical error associated with X-ray beam hardening effects.29 If the finding of Jensen et al.25 is applicable to the current study, it would suggest that the reduction in fat mass could have brought about a loss of almost 50 g of bone mineral or a decrease in total body BMD of ∼2.0%.

These findings regarding weight loss and bone status may affect the interpretation of our results. It is possible that the beneficial effects of exercise, per se, on the skeleton were countered, to some extent, by the negative effects of weight loss, and that the benefits of exercise were therefore underestimated. Whether this may have affected both exercise groups in a similar manner is unknown and likely depends on the mechanism(s) by which weight reduction leads to bone loss. One possibility is that the bone mineral loss that accompanies weight loss is mediated through the reduction in mechanical loading forces secondary to reduced body mass. In this case, it is possible that the beneficial effects of the GRF exercise program were underestimated to a greater extent, since women in that group lost more weight than did those in the JRF exercise program. Alternatively, since adrenal androgens are converted to estrogen in adipose tissue, it may be the decrease in fat mass that leads to bone loss, through a reduction in circulating estrogens. If this is the case, it is likely that both exercise groups were affected similarly, since there were comparable reductions in fat mass. Regardless of the mechanism, the results of the current study indicate that vigorous exercise training can induce significant increases in BMD in older postmenopausal women, despite reductions in body weight and fat mass.

The mechanisms by which exercise induces an increase in BMD remain poorly understood. In young and adult rats with intact ovaries, exercise that results in an increase in bone mass has been shown to have a stimulatory effect on the rate of bone formation.30–33 In ovariectomized rats, however, the increase in bone mass appears to be due to the suppression of bone resorption, as exercise only sustains the high rate of formation that is brought about by the increase in bone turnover subsequent to ovariectomy.34,35 Similarly, it has been shown that serum osteocalcin concentration, which is a marker for bone formation activity, increases in response to exercise training in young women,4 but not in older postmenopausal women.7 It seems possible that in older women, as in ovariectomized rats, the increases in BMD in response to exercise are brought about primarily through inhibition of resorption rather than through the stimulation of the bone formation rate. This is supported in the present study by the observation that exercise of sufficient intensity to increase BMD did not result in an increase in circulating osteocalcin level. In fact, there was a significant reduction in serum osteocalcin in response to the JRF exercise program. However, since this group had a significantly higher baseline serum osteocalcin concentration that was reduced in response to exercise to a level that was similar to those of the other two study groups, we cannot rule out the possibility that this reflected a regression to the mean. Further studies, assessing markers of both bone formation and resorption, are necessary to better describe the mechanisms by which exercise increases BMD in older, postmenopausal women.

The results of this study demonstrate that older, postmenopausal women are able to increase BMD at clinically relevant sites through exercise training programs that introduce stress to the skeleton by means of either ground-reaction (i.e., walking, jogging, stairs) or joint-reaction forces (i.e., weight lifting, rowing). Although the two training programs resulted in similar improvements at several sites, including the whole body, lumbar spine, and Ward's triangle region of the hip, a notable exception was the failure of the JRF exercise program to induce an increase in BMD of the femoral neck. Since low femoral neck BMD is predictive of hip fractures,36 it may seem prudent to recommend exercises that generate ground-reaction forces over those that generate joint-reaction forces for the prevention of osteoporosis. However, some adaptations that were specific to the JRF exercise program, particularly the increases in lean body mass and strength, may be important not only in preventing osteoporotic fractures by reducing the risk for falls,24 but also in preventing the decline in functional independence and overall health with advancing age.37 It remains to be determined whether all of these benefits can be gained through a training program that combines the different types of exercises employed in this study.

Acknowledgements

The authors are grateful for the product support kindly provided by Marion Merrell Dow Inc. and for the excellent technical assistance of Mary Lammert, Kevin Kincaid, Kathie Obert, Debbie Bronder, Karen Krochina, and the staffs of the Applied Physiology Section and the General Clinical Research Center. This research was supported by NIH research awards AR40705 and AG05562, and General Clinical Research Center grant 5-M01-RR00036. Dr. Kohrt was supported by Research Career Development Award AG00663.

Ancillary