Bone Mineral Density Response to Caloric Restriction–Induced Weight Loss or Exercise-Induced Weight Loss: A Randomized Controlled Trial

Background  Bone loss often accompanies weight loss induced by caloric restriction (CR), but whether bone loss accompanies similar weight loss induced by exercise (EX) is unknown. We tested the hypothesis that EX-induced weight loss is associated with less bone loss compared with CR-induced weight loss.

Methods  Forty-eight adults (30 women; 18 men; mean ± SD age, 57 ± 3 years; and mean ± SD body mass index, 27 ± 2 kg/m²) were randomized to 1 of 3 groups for 1 year: CR group (n = 19), regular EX group (n = 19), or a healthy lifestyle (HL) control group (n = 10). Primary outcome measure was change in hip and spine bone mineral density (BMD). Secondary outcomes were bone markers and hormones.

Results  Body weight decreased similarly in the CR and EX groups (10.7% ± 6.3% [−8.2 ± 4.8 kg] vs 8.4% ± 6.3% [−6.7 ± 5.6 kg]; P = .21), whereas weight did not change in the HL group (−1.2% ± 2.5% [−0.9 ± 2.0 kg]). Compared with the HL group, the CR group had decreases in BMD at the total hip (−2.2% ± 3.1% vs 1.2% ± 2.1%; P = .02) and intertrochanter (−2.1% ± 3.4% vs 1.7 ± 2.8%; P = .03). The CR group had a decrease in spine BMD (−2.2% ± 3.3%; P = .009). Despite weight loss, the EX group did not demonstrate a decrease in BMD at any site. Body weight changes correlated with BMD changes in the CR (R = 0.61; P = .007) but not in the EX group. Bone turnover increased in both CR and EX groups.

Conclusions  CR-induced weight loss, but not EX-induced weight loss, is associated with reductions in BMD at clinically important sites of fracture. These data suggest that EX should be an important component of a weight loss program to offset adverse effects of CR on bone.

Trial Registration  clinicaltrials.gov Identifier: NCT00099138

Most adults in the United States are either overweight (body mass index [BMI; calculated as weight in kilograms divided by height in meters squared] 25.0-29.9) or obese (BMI ≥30.0).¹ Lifestyle modification, involving low-calorie diet and exercise (EX), is the primary therapy for overweight and obese persons.² However, body weight also affects bone mineral density (BMD), the strongest predictor of future osteoporotic fracture.³ Increased body weight is associated with decreased osteoporosis and hip fractures in older men and women, whereas the converse is true for decreased body weight.^4-8 The mechanisms for the osteoprotective effect of body weight are not completely understood but are likely to include increased skeletal loading and increased plasma levels of bone-active hormones (eg, estradiol).⁹

Data from prospective intervention studies have demonstrated that weight loss often results in bone loss and bone resorption.^10-20 However, the clinical importance of these findings are unclear because most studies were conducted in obese persons, who usually have increased BMD, were of short duration (≤6 months) and reported whole-body BMD rather than regional BMD at the clinically important sites of fracture. Moreover, in almost all studies, weight loss was induced by caloric restriction (CR) rather than by increasing energy expenditure through EX, which is considered a beneficial therapy for osteoporosis.²¹ Although EX without increase in energy intake causes weight loss,²² little is known about the effects of EX-induced weight loss on BMD and bone turnover.

The purpose of the present study was to conduct a 12-month, randomized controlled trial to test the hypothesis that weight loss induced by EX results in less bone loss compared with weight loss induced by CR at clinically important sites of osteoporotic fracture (ie, hip and spine) in nonobese middle-aged men and women. The data reported in this article were obtained as part of an investigation of the feasibility of CR in healthy volunteers (Comprehensive Assessment of Long-term Effects of Reducing Intake of Energy [CALERIE]).²³

This study was conducted at Washington University School of Medicine and was approved by the Human Studies Committee and the General Clinical Research Center Advisory Committee, St Louis, Mo. Written informed consent was obtained from each subject. Healthy men and women were recruited by using local advertisements. Eligible participants were 50 to 60 years old with a BMI between 23.5 and 29.9 (ie, nonobese), held a stable weight (<2% change in body weight) for at least 3 months, and were nonsmokers; women had to be postmenopausal. Exclusion criteria included regular EX (>20 minutes of EX more than twice per week), history or clinical evidence of cardiopulmonary disease, history of cancer, and use of bone-acting drugs (eg, bisphosphonates and glucocorticoids) within the previous year. Menopausal hormone therapy was not exclusionary, but women receiving hormone therapy had to be taking a stable dose for at least 6 months before enrollment into the study. The effects of CR on body composition and glucose tolerance in these subjects were reported recently.^23,24

After screening evaluations, 48 participants were randomized, with stratification for sex, to 1 of the following 3 groups in a 2:2:1 sequence: CR group (n = 19), EX group (n = 19), or healthy lifestyle (HL) control group (n = 10) for 12 months. More participants were assigned to the intervention groups to increase the number of participants in whom CR and EX could be evaluated. The results of recruitment and randomization have been previously described.²³

The CR intervention goal was to decrease energy intake by 16% during the first 3 months and by 20% during the remaining 9 months. Diet prescriptions were based on baseline energy intake, which was assumed to be equal to total daily energy expenditure as determined by the doubly labeled water method. Participants met individually with dietitians weekly to determine strategies for reducing energy intake and attended weekly group meetings. Further details about the CR intervention, including compliance data, have been reported previously.^23,24

The EX intervention goal was to induce the same calorie deficit as was induced by the CR intervention by maintaining energy intake at baseline and increasing energy expenditure by 16% of baseline total daily energy expenditure for the first 3 months and by 20% for the subsequent 9 months. Exercise trainers worked closely with the participants to monitor their EX energy expenditure goals. The participants exercised while using heart rate monitors (Polar S610; Polar Electro Oy, Kempele, Finland). Additional details about the EX intervention, including compliance data, have been reported previously.^23,24

The HL control group did not receive advice to change diet or EX habits. They were offered information about a healthy diet²⁵ but only received it if requested. Participants in all groups were given a multivitamin with mineral supplement (162 mg of calcium and 400 IU of cholecalciferol) (Sentro-Car; Darby Co, Reno, Nev).

Body weight was measured at baseline and 1, 3, 6, 9, and 12 months. Body weight was measured in duplicate in the morning following a 12-hour fast, with the participant wearing a hospital gown without shoes. Bone mineral density of the total body, lumbar spine, and proximal femur was measured every 3 months by using dual-energy x-ray absorptiometry (Delphi 4500-W; Hologic Corporation, Waltham, Mass); BMD of the lumbar spine was performed using the anteroposterior projection and was calculated as the mean BMD of vertebrae L1-L4. The coefficient of variation for this technique at our center is 1.1% for the lumbar spine and 1.2% for the proximal femur.²⁶

Dietary intakes were determined by using 7-day food diaries, which were analyzed with Nutrition Data System for Research (Minneapolis, Minn). Physical activity was determined by using the Stanford 7-Day Physical Activity Re-call Questionnaire,²⁷ which quantifies the time engaged in activities and the metabolic equivalents (METs) for each activity. Physical Activity Recall Questionnaire data are presented as MET hours per day above rest, as reported previously.²³

A venous blood sample was taken in the morning after subjects fasted for at least 12 hours at baseline and at 6 and 12 months. Enzyme-linked immunosorbent assay kits were used to measure C-telopeptide of type I collagen (Crosslaps; Nordic Bioscience Diagnostics, Herlev, Denmark; coefficient of variation, 2.1%) as a marker of bone resorption and bone-specific alkaline phosphatase (Metra BAP; Quidel Corporation, San Diego, Calif; coefficient of variation, 4.9%) and osteocalcin (Metra OC; Quidel Corporation, San Diego coefficient of variation, 4.4%) as markers of bone formation. Radioimmunoassay kits were used to measure estradiol (Ultra-sensitive estradiol DSL-4800; Diagnostic Systems Laboratories Inc, Webster, Tex; coefficient of variation, 7.5%) and leptin (Leptin HL-81K; Linco Research Inc, St Charles, Mo; coefficient of variation, 5.6%).

All subjects who provided follow-up data at any time point are included in the analysis. Between-group comparisons of categorical baseline variables were performed using χ² tests or Fisher exact tests. Between-group comparisons of continuous baseline variables were performed using analysis of variance.

Longitudinal analyses were carried out using mixed-model repeated-measures analysis of variance. The primary focus of these analyses was on the significance of the interaction between group and time point. Within the framework of the mixed model, when the P value for an interaction was ≤.10, the appropriate contrast was used to test the null hypothesis that changes between 2 specific time points in 1 group were equal to corresponding changes in another group. Analyses testing for within-group changes also were performed using mixed-model repeated-measures analysis of variance. Select mixed models were adjusted for sex, age, and current hormone therapy use depending on the outcomes examined. Data are presented in tables as least squares mean ± SE. Data are presented in text as mean ± SD. All statistical tests were 2-tailed, and P≤.05 was considered statistically significant. The data analysis was generated using SAS statistical software (version 9.1.3; SAS Institute, Cary, NC).

Of the 48 subjects who started the intervention, 46 completed the study; 1 woman dropped out of the CR group at 6 months because of the inability to comply with the diet prescription, and 1 man dropped out of the EX group at 9 months because of medical reasons unrelated to the study.

The 3 groups did not differ on baseline demographic and clinical characteristics except for age, with the EX group being slightly older than the CR group (Table 1). Most participants were overweight (BMI, 27.3 ± 2.0). The proportion of women receiving hormone therapy did not differ among the groups. One woman in the HL group and 1 woman in the EX group discontinued hormone therapy after the 3-month time point. Therefore, for these 2 participants, follow-up data collected after 3 months were not used.

Body weight decreased similarly in the CR group (−10.7% ± 6.3%; −8.2 ± 4.8 kg) and EX group (−8.4% ± 6.3%; −6.7 ± 5.6 kg), whereas body weight did not change in the HL group (−1.2% ± 2.5%; −0.9 ± 2.0 kg) (Table 2). The decrease in body weight in the CR group was accompanied by significant decreases in regional BMD, specifically at the lumbar spine (−2.2% ± 3.3%), total hip (−2.2% ± 3.1%), and intertrochanter (−2.1% ± 3.4%). The changes in BMD at the total hip and intertrochanter were significantly different from the corresponding changes (total hip: 1.2% ± 2.1%; intertrochanter: 1.7% ± 2.8%) in the HL group (Table 2). In contrast, despite the decrease in body weight in the EX group, there were no significant changes in BMD in this group, and there were no significant differences in BMD changes between the EX and HL groups. Accordingly, changes in body weight correlated with changes in hip BMD in the CR group but not in the EX group (P = .02 for difference between correlations; Figure 1).

Bone markers and hormones over the study period are given in Table 3 and Figure 2. C-telopeptide of type I collagen levels significantly increased in the CR group and EX group at 6 months, while C-telopeptide of type I collagen levels did not change in the HL group. Although there were no significant changes in bone-specific alkaline phosphatase levels from baseline within each group, the change in bone-specific alkaline phosphatase at 6 months in the EX group was greater than the corresponding changes in the CR and HL groups. There were no significant changes in osteocalcin concentrations. Leptin concentrations decreased in the CR and EX groups compared with no changes in the HL group. Changes in leptin concentrations did not correlate with changes in BMD (all P>.10). There were no significant changes in serum estradiol levels.

Seven-day food records showed that the CR group restricted their energy intake (−382 ± 404 kcal/d), while the EX (+9.6 ± 213 kcal/d) and HL (+67 ± 237 kcal/d) groups maintained their intake (Table 4). Accordingly, intake of macronutrients such as protein and fat also decreased in the CR compared with the EX and HL group. The intake of many micronutrients increased in the EX and CR groups, although not in the HL group.

Seven-day Physical Activity Recall Questionnaire findings indicated that MET hours per day increased in the EX group (from 10.4 ± 2.5 to 14.1 ± 4.4) but remained unchanged in the CR (from 11.0 ± 2.9 to 10.1 ± 1.3) and HL (from 9.8 ± 1.9 to 10.9 ± 3.4) groups (P value between groups, .01). Heart rate monitors revealed that the EX group exercised 5.8 ± 2.5 sessions per week for 62.5 ± 17.8 minutes per session, at an intensity of 72% ± 9% of maximal heart rate and expended 317 ± 160 kcal/d. The most frequently used modes of EX were walking and/or jogging, elliptical machines, and cycle ergometers. Walking and/or jogging was reported 2.7 times more frequently than elliptical training and 5 times more frequently than cycling.

Maintaining adequate bone mass helps reduce fracture risk in old age. Although bone loss often accompanies CR-induced weight loss in obese persons,^10-18 the clinical implications of this adverse effect are unclear because obesity is associated with increased bone mass. In addition, it is unknown whether EX-induced weight loss causes the same decrease in bone mass as CR-induced weight loss. Therefore, we conducted a 1-year randomized controlled trial to evaluate the effect of similar weight loss induced by either CR or EX on BMD in nonobese, middle-aged adults. The results of the present study show that EX-induced weight loss does not change BMD, whereas CR-induced weight loss decreases BMD in the lumbar spine, total hip, femoral neck, and intertrochanter. Changes in body weight correlated directly with changes in BMD in the CR group but not in the EX group. These findings have important implications in designing an appropriate weight-loss therapy program in middle-aged adults, particularly in the subset of patients who may already be at increased risk for bone fracture.

Previous studies have shown that CR causes bone loss that may be proportional to the amount of weight lost; the approximate 2% loss of hip BMD in the CR group is consistent with the observed weight loss of approximately 10%.^10-20 On the other hand, previous studies have shown that EX generally has positive effects on BMD.²¹ However, in these studies, EX usually resulted in little or no weight loss because of increased energy intake. Thus, there is little information on the effects of EX-induced weight loss on BMD. To our knowledge, only 1 previous study compared the effects of CR- and EX-induced weight loss on BMD.²⁸ In that study, weight loss was 2-fold greater in the CR group (approximately 6%) than in the EX group (approximately 3%), and the CR group lost BMD (1.5%), whereas the EX group did not. In the present study, we carefully monitored energy balance (eg, using food diaries and heart rate monitors) to (1) reduce caloric intake without changing energy expenditure in the CR group and (2) increase energy expenditure without changing caloric intake in the EX group. Our study, therefore, provides novel information on the skeletal effects of CR- and EX-induced weight loss. Despite both weight loss and increased bone turnover in the CR and EX groups, only the CR group had a significant decrease in BMD.

A common explanation given for the bone loss induced by weight loss is reduction in mechanical stress on the weight-bearing skeleton (ie, hip and spine).^13,29 Accordingly, the preservation of BMD in the EX group could be mediated through EX-induced bone loading.³⁰ However, because bone loss with weight loss also occurs in nonweight bearing sites,³¹ other mechanisms for weight loss–induced bone loss have been proposed, such as the decline in bone-active hormones produced by adipocytes.³² Among such hormones, leptin has emerged as a candidate responsible for the protective effects of fat on bone³³ and has been shown to decrease receptor activator of nuclear factor κB ligand levels³⁴ and increase osteoprotegerin levels with inhibition of osteoclastogenesis.³⁵

As expected with weight loss, we found that leptin levels decreased similarly in the EX and CR groups. Accordingly, fat mass decreased to a similar extent (approximately 6 kg) in the EX and CR groups, as previously reported.²³ However, leptin changes did not correlate with BMD changes, and the EX group did not have a decrease in BMD despite reduced leptin concentrations. Therefore, our results do not support an important role for leptin. There were also no significant changes in estradiol levels, despite use of an ultrasensitive estradiol assay. However, these findings must be interpreted cautiously because approximately one third of female subjects were receiving hormone therapy (although we enrolled only women using a stable dose prior to the study). These findings are consistent with previous reports demonstrating that (1) the association with BMD is stronger for weight than for estrogen concentrations⁹ and (2) estrogen production may not be the most significant predictor of BMD.³⁶

There is little information regarding markers of bone turnover during weight loss,^10,20,37 and to our knowledge the present study is the first to measure C-telopeptide of type I collagen, a sensitive marker of bone resorption.³⁸ As indicated by the increase in C-telopeptide of type I collagen levels, bone turnover was stimulated in both the CR and EX groups. However, the increase in bone turnover was detrimental only in the CR group, while the EX group did not have a significant decrease in BMD even after 1 year. The changes in bone-specific alkaline phosphatase levels (marker of bone formation) in the EX group were greater than in the CR group, suggesting less imbalance in the bone remodeling cycle.³⁹ Exercise is thought to act on the skeleton through muscle pull, producing strains in the skeleton that are perceived by bone cells as osteogenic.⁴⁰ Our results are consistent with an osteoprotective effect of EX-induced mechanical strain on the skeleton and consequent increase in bone turnover.^17,41

The strengths of the present study include the randomized controlled trial design, the comprehensive assessments of energy intake and expenditure, and the high rate of compliance of participants. We used dual-energy x-ray absorptiometry to monitor bone loss because it is the standard method of measuring BMD (accounting for ≥60% of the variance in bone strength) in clinical and research settings. A limitation is that the dual-energy x-ray absorptiometry provided information on bone quantity but not bone quality (eg, microarchitecture), which is an additional determinant of fracture risk.⁴² Further studies are needed to evaluate bone quality in response to CR- vs EX-induced weight loss. Because the present study was part of a parent study of the feasibility of CR in humans and effects on body composition,²³ an additional limitation is the relatively small sample size. Our findings must be considered preliminary, particularly as it was not powered to examine sex differences in BMD responses. We controlled for the effect of sex by including it as a covariate in the mixed-model analysis of variance. Most participants were overweight, so the results should be cautiously extended to nonoverweight populations.

In summary, 1 year of CR and 1 year of EX induced similar weight loss, but only CR-induced weight loss was accompanied by significant decreases in BMD. Our results provide evidence that EX-induced weight loss is associated with preservation of BMD at important clinical sites of fracture. Therefore, EX has the important advantage over CR by protecting against bone loss. However, because the amount of EX required to achieve clinically meaningful weight loss is large, a more practical approach for weight reduction is a combination of CR and EX. Our results suggest that regular EX should be included as part of a comprehensive weight loss program to offset the adverse effects of CR on bone.

Correspondence: Dennis T. Villareal, MD, 4488 Forest Park Blvd, Washington University School of Medicine, St Louis, MO 63110 (dvillare@wustl.edu).

Accepted for Publication: September 14, 2006.

Author Contributions: Dr Villareal had full access to all the data in the study and takes responsibility for the integrity and the accuracy of the data analysis. Study concept and design: Holloszy and Schechtman. Acquisition of data: Villareal and Weiss. Analysis and interpretation of data: Villareal, Fontana, Weiss, Racette, Steger-May, Schechtman, Klein, and Holloszy. Drafting of the manuscript: Villareal, Fontana, Racette, Steger-May, Schechtman, and Klein. Critical revision of the manuscript for important intellectual content: Villareal, Fontana, Weiss, Racette, Schechtman, Klein, and Holloszy. Statistical analysis: Steger-May and Schechtman. Obtained funding: Holloszy. Administrative, technical, and material support: Holloszy, Villareal, Fontana, Racette, and Klein. Study supervision: Villareal, Weiss, and Racette.

Financial Disclosure: None reported.

Funding/Support: This work was supported by Cooperative Agreement grant AG20487, General Clinical Research Center grants RR00036 and RR02602, Diabetes Research Training Center grant DK20579, and Clinical Nutrition Research grant Unit DK56341 from the National Institutes of Health. Dr Weiss was supported by grant AG00078 from the National Institutes of Health.