Sex Differences in the Effects of Weight Loss Diets on Bone Mineral Density and Body Composition: POUNDS LOST Trial

Abstract

Context:

Weight loss is associated with reduction in bone mineral density (BMD).

Objective:

The objective was to address the role of changes in fat mass (FM) and lean mass (LM) in BMD decline in both sexes.

Design:

A 2-year randomized controlled trial, the Preventing Overweight Using Novel Dietary Strategies (POUNDS-LOST).

Setting:

The setting was the general community.

Patients or Other Participants:

Enrolled were 424 overweight and obese participants (mean age, 52 ± 9 y; 57% females).

Intervention:

Intervention included weight loss diets differing in fat, protein, and carbohydrates.

Main Outcome Measures:

Main outcome measures were change in spine, total hip (TH), and femoral neck (FN) BMD and sex differences after dietary intervention.

Results:

At baseline, a stronger correlation between BMD and body composition measurements was observed in women, primarily with LM (r = 0.419, 0.507, and 0.523 for spine, FN, and TH, respectively; all P < .001). In men, only LM correlated with hip BMD (r = 0.298; P < .001). Mean weight loss at 2 years was −6.9%, without differences among diets. Two-year changes in BMD were 0.005 (P = .04), −0.014 (P < .001), and −0.014 g/cm² (P < .001), at the spine, TH, and FN, respectively. These changes directly correlated with changes in LM in women (r = 0.200, 0.324, and 0.260 for spine, FN, and TH, respectively), whereas FM loss correlated only with changes in TH BMD (0.274; P < .001). In men, changes in LM (−0.323; P < .001) and FM (−0.213; P = .027) negatively correlated with changes in spine BMD.

Conclusions:

Weight loss diets result in sex-specific effects on BMD. Although men exhibited a paradoxical increase in spine BMD, women tended to decrease in BMD at all sites.

With the epidemic of overweight and obesity worldwide, there is a growing interest in diets that are efficacious for weight loss, without adversely affecting other aspects of overall health, such as bone mineral density (BMD), body composition, or nutritional adequacy. In recent large randomized controlled trials, both low-carbohydrate and low-fat diets resulted in a comparable degree of weight loss without significant adverse effects of any of the dietary regimens (1, 2). Although the effects of the macronutrient composition on cardiovascular risk factors were studied in relatively large cohorts with long follow-up, the potential adverse effects of weight loss per se and of the different dietary regimens on bone health are less clear.

Some, but not all, studies show that weight loss diets are associated with clinically important decreases in BMD and increases in markers of bone turnover, whether achieved through dietary interventions (2–7) or bariatric surgery (8–10). However, most studies on bone loss with calorie-restricted diets were small and were conducted primarily in women, thereby limiting their generalizability. A randomized controlled trial comparing low-carbohydrate to low-fat diets did not reveal any between-treatment differences in weight loss or BMD changes after 2 years of intervention (11), but other studies suggested that high protein intake (especially animal protein) may lead to decreased BMD (12, 13). In the prospective Study of Osteoporotic Fractures Research cohort of over 1000 women, those with a high dietary ratio of animal to vegetable protein intake were found to have more rapid femoral neck (FN) bone loss (14). In contrast, higher protein intake was inversely associated with forearm fractures and was associated with improved total and hip BMD values (15). Although previous studies show positive associations between lean mass (LM), fat mass (FM), and BMD, there are few data on the relationship between longitudinal changes in body composition and BMD during weight loss in women and men (16).

The Preventing Overweight Using Novel Dietary Strategies (POUNDS LOST) trial (www.clinicaltrials.gov; registration no. NCT00072995) is a randomized, controlled trial that examined the effects of four energy-restricted diets varying in fat, protein, and carbohydrate content on weight loss in 811 overweight and obese women and men. In the current study, we investigated the following among POUNDS LOST participants in whom BMD and body composition measurements were available: 1) the prospective effects of the dietary interventions and weight loss on changes in BMD in both women and men; 2) the correlations between changes in BMD and body composition (fat and lean tissue) in both sexes; and 3) whether there are sex differences in the relationships between changes in BMD and body composition measures.

Subjects and Methods

Study design and sites

The study was designed as a randomized clinical trial to compare effects on body weight of energy-reduced diets that differed in intended macronutrient composition—low (20%) or high (40%) in fat, average (15%) or high (25%) in protein, and across a range of carbohydrate (from 35 to 65%), as previously described (1). The trial was conducted from October 2004 through December 2007 at two sites: the Harvard School of Public Health and Brigham and Women's Hospital (BWH), Boston, Massachusetts; and the Pennington Biomedical Research Center of the Louisiana State University System, Baton Rouge, Louisiana.

Participants

Of the 811 participants in the dietary intervention trial, 424 participants (242 women and 182 men) were enrolled in the BMD and body composition study. BMD and fat and lean tissue masses were measured at baseline and after 6 months and 2 years of the dietary interventions. Participants were 30 to 70 years of age (mean age, 51.8 ± 8.9 y) with a body mass index (BMI; weight in kilograms divided by the square of the height in meters) of 25 to 40 kg/m². Major criteria for exclusion were the presence of diabetes or unstable cardiovascular disease, the use of medications that affect body weight, and insufficient motivation as assessed by interview and questionnaire. Menopausal status was determined at enrollment and was self-reported based on the absence of menstrual periods during the previous 12 months or surgical removal of both ovaries. The study was approved by the Institutional Review Board at each institution and by a data and safety monitoring board appointed by the National Heart, Lung, and Blood Institute. All participants provided written informed consent. Random assignments to one of four diet groups were generated by the data manager at the coordinating center on request of a study dietitian, after the eligibility of a participant was confirmed.

Weight loss intervention

The nutrient compositions of the four diets at 6 months of intervention (the maximal weight loss) were: 26.2% fat, 17.6% protein, and 57.5% carbohydrates (low-fat, average-protein); 25.9% fat, 21.8% protein, and 53.4% carbohydrates (low-fat, high-protein); 33.9% fat, 18.4% protein, and 49.1% carbohydrates (high-fat, average-protein); and 34.3% fat, 22.6% protein, and 43% carbohydrates (high-fat, high-protein). Thus, two diets were average-protein (17.6 and 18.4% of total calories) and two were high-protein (21.8 and 22.6% of total calories). The four diets also allowed for a dose–response test of carbohydrate intake ranging from 43 to 57.5% of energy. Blinding was maintained by using similar foods for each diet. Investigators and staff who measured outcomes were unaware of the participants' diet assignment. All participants joined an intensive lifestyle program, including a goal for physical activity for 90 minutes of moderate exercise per week. Participation in exercise was monitored by questionnaire and by the online self-monitoring tool.

Measurements

Body weight and waist circumference were measured in the morning before breakfast on 2 days at baseline, 6 months, and 2 years. Dietary intake was assessed in a random sample of 50% of participants by reviewing the 5-day diet record at baseline and by 24-hour recall at 6 and 24 months. BMD of the spine (L1–L4), FN, and total hip (TH) and body composition were measured after an overnight fast using dual-energy x-ray absorptiometry (DXA) (Hologic QDR-4500A bone densitometer; Hologic, Inc) at both study centers, with the participant in the supine position and wearing a hospital gown at baseline, 6, and 24 months. Results are expressed as BMD (g/cm²). Body composition values were analyzed using APEX software version 3.3 (Hologic) that allows calculation of adipose and LM and rate of change reports for fat and lean tissue. Changes in total adiposity were determined by total body fat, percentage total fat, and FM index (FM/h²; FMI), and changes in LM included lean body mass and the LM index (LM/h²; LMI) (17). In vivo precisions (coefficient of variation) for spine and FN measurements on different days at BWH were 0.68 and 0.99% and 1.21 and 1.74%, for premenopausal and postmenopausal women, respectively (18, 19). Coefficient of variation for fat and lean tissue measures were 1.09 ± 0.15% and 0.89 ± 0.28% (19). Visceral adipose tissue (VAT) was assessed by quantitative computed tomography (20). Dietary intake of calcium was calculated from the nutritional evaluation of the POUNDS LOST participants using a 5-day diet record at baseline and by 24-hour recall at 6 and 24 months. We do not have information on the use of supplemental calcium or vitamin D. Serum calcium and 25-hydroxyvitamin D levels were not measured as part of the POUNDS LOST study.

Digital files from BWH were reanalyzed by a single reader at the Pennington Biomedical Research Center. Daily phantoms were used to ensure instrument stability over time. A three-point body fat phantom was used to verify the accuracy of the two instruments across study sites; no correction was applied because the two instruments were well matched and stable over time.

Statistical analysis

Analyses were performed with SAS software version 9.3 (SAS Institute). The level of significance for all tests was set to P < .05. Baseline data are reported as means ± SD for continuous variables and as counts (percentage) for nominal variables. Changes from baseline data are expressed as means ± SEM. Differences between sexes (men and pre- and postmenopausal women) were assessed using one-way ANOVA for continuous variables and χ² or Fisher's exact tests for nominal variables. Pearson's correlation coefficients were computed to examine relationships between baseline measurements of BMD or bone mineral content and changes in these measures over time.

Models

Diet and sex effects

The primary outcome of the study was the change in BMD over a period of 2 years of dietary intervention in both women and men. To determine whether changes over time differed by sex or macronutrient composition, the generalized linear model ANOVA was used with change scores (6-mo or 2-y value minus the baseline value) as dependent variables. Main effects in these models were protein amount (15 or 25%) and fat amount (20 or 40%). We included baseline values, site, and age as covariates in the models. Prespecified comparisons of interest were high-protein (25%) compared with average-protein (15%) diets, high-fat (40%) compared with low-fat (20%) diets, and high-carbohydrate (65%) compared with low-carbohydrate (35%) diets, and between men and women (at both age groups). When the generalized linear model ANOVA identified a between-groups difference, we determined which pairs of groups differed by using the Tukey-Kramer method. There were no significant interactions between protein and fat contents in any of these models, which justified the analysis by factorial design.

Tests for trend across carbohydrate assignment and intake

We tested for a trend across the carbohydrate amount assigned (35–65%) by using multiple linear regression. In these models, change score (6-mo or 2-y value) was the dependent variable, carbohydrate amount was the main continuous predictor, and baseline value, age, and site were covariates.

Associations between changes in body weight and composition and BMD changes

The strength and direction of associations between weight change, as well as changes in LM, LM index (LMI), FM, FM index (FMI), fat to LM ratio, and VAT and change in BMD at the various sites were assessed by Pearson correlation coefficients. Stepwise multivariable regression was used in an exploratory fashion to identify predictors of the change in total BMD.

Results

At baseline, DXA measurements were obtained on 424 participants (242 women and 182 men), and 328 (77.3%) and 236 (55.7%) study participants completed the study at 6 and 24 months, respectively. Table 1 shows the baseline characteristics of the study participants. Mean age at baseline was 51.8 ± 8.9 years. Mean BMI was 32.6 ± 3.8 kg/m², with no significant differences among dietary intervention groups. Baseline BMD at the spine, TH, and FN was similar between assigned diet groups. Although there were differences in the carbohydrate and fat composition among the four diets, there were no differences in total energy intake across groups.

Table 1.

Baseline Characteristics of the Study Population

Diet (Prescribed)	35% Carb	45% Carb	55% Carb	65% Carb	All	P Valuea,b
% Carbohydrates	35	45	55	65
% Protein	25	15	25	15
% Fat	40	40	20	20
n	102	107	108	107	424
Age, y	52.2 ± 8.6	52.1 ± 9.3	52.1 ± 9.3	51.0 ± 8.6	51.8 ± 8.9	.738
Sex (F/M), n	58/44	58/49	66/42	60/47	242/182	.769
BMD data
Vertebral	1.039 ± 0.146	1.058 ± 0.140	1.068 ± 0.144	1.070 ± 0.146	1.059 ± 0.144	.404
TH	1.028 ± 0.152	1.017 ± 0.136	1.006 ± 0.125	1.020 ± 0.128	1.018 ± 0.135	.706
FN	0.852 ± 0.146	0.836 ± 0.123	0.841 ± 0.126	0.847 ± 0.118	0.844 ± 0.128	.827
Anthropometric measurements
Height, cm	170.0 ± 9.5	168.9 ± 9.2	168.1 ± 8.6	169.6 ± 8.5	169.1 ± 8.9	.427
Weight, kg	95.8 ± 17.8	93.6 ± 16.1	91.5 ± 13.0	94.4 ± 15.7	93.8 ± 15.7	.244
BMI, kg/m2	33.0 ± 4.2	32.6 ± 3.8	32.3 ± 3.6	32.7 ± 3.7	32.6 ± 3.8	.719
Waist circumference, cm	105.5 ± 13.5	104.8 ± 13.4	103.4 ± 11.3	104.5 ± 13.6	104.5 ± 13.0	.704
Body fat, %	37.3 ± 6.7	37.1 ± 7.1	37.6 ± 6.8	36.5 ± 6.7	37.1 ± 6.8	.678
FM, kg	35.5 ± 8.3	34.5 ± 7.6	34.2 ± 7.5	34.3 ± 8.3	34.6 ± 7.9	.619
LM, kg	60.3 ± 14.2	59.1 ± 13.7	57.3 ± 11.0	60.1 ± 12.5	59.2 ± 12.9	.295
VATc	5.0 ± 2.5	5.4 ± 2.6	5.7 ± 2.5	5.6 ± 2.5	5.4 ± 2.5	.548
Subcutaneous adipose tissued	10.6 ± 2.6	11.6 ± 2.5	11.0 ± 2.8	10.9 ± 2.5	11.1 ± 2.6	.363
Female hormone use, n	58	58	66	60	242
Hormone replacement therapye	12 (27.9%)	11 (28.2%)	17 (41.5%)	15 (50.0%)	55 (35.9%)	.778
Oral contraceptivesf	3 (20.0%)	0 (0.0%)	4 (16.0%)	5 (16.7%)	12 (13.5%)	.142
Dietary datag
Energy, kcal	2025 ± 598	2027 ± 602	1879 ± 581	2071 ± 496	2000 ± 573	.122
% Protein	18 ± 3	18 ± 3	18 ± 4	18 ± 3	18 ± 3	.872
% Carbohydrates	44 ± 7a	44 ± 7	47 ± 8b	44 ± 8	45 ± 8	.035
% Fat	38 ± 6a	37 ± 5	36 ± 6b	38 ± 6	37 ± 6	.027
Alcohol, g	6 ± 9	7 ± 9	4 ± 7	6 ± 8	6 ± 8	.147
Smoking (current/past/never), %	5/42/53	4/41/55	2/25/73	4/43/53	4/38/59	.033
Physical activity ratioh	1.58 ± 0.11	1.58 ± 0.10	1.57 ± 0.11	1.58 ± 0.11	1.58 ± 0.11	.956
Dietary calcium, mg/d	901 ± 561	959 ± 486	867 ± 402	971 ± 605	925 ± 519	.421
Urinary nitrogen,g	12.9 ± 5.0	12.0 ± 4.3	11.7 ± 4.1	13.0 ± 3.8	12.3 ± 4.3	.065

Abbreviations: Carb, carbohydrate; F, female; M, male. Data are expressed as mean ± SD, unless specified otherwise. Bold indicates significant values (P < .05).

^a,bValues are significantly different from one another at P < .10 after using Scheffe's adjustment for multiple comparisons.

^cn = 194.

^dn = 165.

^en = 153 (43 on 35%; 39 on 45%; 41 on 55%; 30 on 65%).

^fn = 89 (15 on 35%; 19 on 45%; 25 on 55%; 30 on 65%).

^gBaseline data are from the average of five diet records available for n = 359 at baseline (85%).

^hPhysical activity was estimated using the Baecke Questionable.

Average weight loss after 6 and 24 months of intervention was 7.9 and 6.4% of initial body weight, respectively, with no significant differences among diet groups (P = .81, 6 mo; P = .34, 24 mo, data not shown). Overall, the change from baseline in spine, FN, and TH BMD values at 24 months including females and males was 0.005, −0.013, and −0.014 g/cm², respectively (all P < .02), with no significant differences among dietary groups (Table 2). In a multivariate model adjusted for age, baseline BMD, diet group (categorical), change in physical activity (0–24 months), baseline height and weight, for every 1% of body weight loss, there was a small decrease in both spine (0.0013 g/cm²; r² = 0.176; P = .002) and FN (0.002 g/cm²; r² = 0.298; P < .001) BMD values in women, but not in men (−0.0009 g/cm²; r² = 0.07; P = .032 for spine BMD; and 0.0005 g/cm²; r² = 0.08; P = .214 for FN BMD).

Table 2.

Changes in BMD During 2 Years of Dietary Intervention and Comparison Between Diet Groups

BMD, g/cm2	Δ0–6 Months	Δ0–24 Months	Time Effect (Continuous)	Comparisons Over 2 Years Between the Composition of Macronutrients in the Various Dietary Regimens
				HiP vs AvP	HiF vs LoF	HiC vs LoC	P Value of Trend
Spine	0.001 ± 0.002	0.005 ± 0.002	0.0116	0.4215	0.1242	0.6507	.3089
FN	−0.013 ± 0.002	−0.013 ± 0.002	<0.0001	0.8026	0.4672	0.4932	.4448
TH	−0.009 ± 0.001	−0.014 ± 0.002	<0.0001	0.3446	0.4402	0.9219	.7902

Abbreviations: HiP, high-protein diet; AvP, average protein diet; HiF, high-fat diet; LoF, low-fat diet; HiC, high-carbohydrate diet; LoC, low-carbohydrate diet. Data were adjusted to baseline parameters. Δ0–6 months and Δ0–24 months represent the changes over baseline at 6 and 24 months of follow-up, respectively. The changes in BMD at the spine, FN, and TH were compared between diets based on the dietary composition of protein, fat, and carbohydrates.

Overall, 52 of 162 (32.1%) postmenopausal women were taking hormone replacement therapy, and 12 of 80 (15%) premenopausal women were taking oral contraceptives. Information on adherence rates of the various hormone regimens was not available to us. After 24 months of dietary intervention, postmenopausal women had significant bone loss at both the spine and FN, with a larger loss at the FN. Premenopausal women exhibited decreased BMD at the FN only, with no change in FN BMD in men. Unexpectedly, at the end of the study, men exhibited an increase in spine BMD (+0.022 ± 0.003 g/cm²; P < .005).

Given the differences in BMD response of women and men to the dietary intervention, we next assessed the changes in body composition (Table 3). The reduction in FM relative to LM was greater in males, resulting in sex differences in fat-to-lean mass ratio at 2 years: 0.51 ± 0.01 in men, and 0.57 ± 0.01 and 0.58 ± 0.02 in pre- and postmenopausal women, respectively (P < .001 for both comparisons vs males). In addition, reduction in VAT in postmenopausal women was associated with significant reductions in BMD at the spine and hip, with men showing a trend toward an inverse association between changes in VAT and spinal BMD.

Table 3.

Changes in Body Composition and BMD After 2 Years of Dietary Intervention, Adjusted for Baseline Values, Diet Assignment, Change in Physical Activity, and Age

Variable	Postmenopausal Women (n = 100 to 134)		Premenopausal Women (n = 29 to 60)a		Men (n = 107 to 154)		Sex Difference (P Values)a			Sex Effect	Sex × BL
	Δ0–24	P Value	Δ0–24	P Value	Δ0–24	P Value	Post vs Pre	Post vs Men	Pre vs Men
Weight, kg	−4.7 ± 0.9	<.0001	−4.2 ± 1.1	.0002	−5.2 ± 0.8	<.0001	1.000	1.000	1.000	0.711	0.620
LM, kg	−2.0 ± 0.6	.002	−2.6 ± 0.7	.0001	−1.3 ± 0.5	.01	1.000	1.000	.352	0.265	0.381
FM, kg	−3.0 ± 0.5	<.0001	−3.1 ± 1.1	.005	−5.9 ± 0.5	<.0001	1.000	<.001	.059	0.331	0.054
LMI	−0.69 ± 0.15	<.0001	−0.95 ± 0.19	<.0001	−0.56 ± 0.15	.0002	.857	1.000	.303	0.544	0.589
FMI	−1.07 ± 0.20	<.0001	−1.06 ± 0.42	.011	−2.16 ± 0.21	<.0001	1.000	.0007	.056	0.349	0.060
VAT, kga	−0.56 ± 0.25	.025	n/a	n/a	−0.71 ± 0.34	.034	n/a	.700	n/a	0.095	0.032
SAT, kga	−1.15 ± 0.26	<.0001	n/a	n/a	−1.78 ± 0.30	<.0001	n/a	.123	n/a	0.534	0.314
Fat/lean ratio at 24 mo	0.57 ± 0.01	<.0001	0.58 ± 0.02	<.0001	0.51 ± 0.01	<.0001	1.000	.003	.013	0.663	0.194
Spine BMD	−0.010 ± 0.004	.005	−0.016 ± 0.007	.040	+0.022 ± 0.003	<.0001	1.000	<.0001	<.0001	0.288	0.205
Femoral neck BMD	−0.033 ± 0.004	<.0001	−0.024 ± 0.008	.002	+0.0008 ± 0.003	.816	.948	<.0001	.009	0.046	0.015

Abbreviations: Pre, premenopausal women; Post, postmenopausal women; n/a, not available; SAT, sc adipose tissue; BL, baseline. Menopausal status was self-reported at enrollment.

^aFor computed tomography measures (VAT and SAT), n = 42 postmenopausal women, 0 premenopausal women, and 55 men.

Correlation between body composition and BMD changes

The associations between 2-year changes in LMI and FMI and BMD at the spine and hip are shown in Table 4. The change in LMI in postmenopausal women and men produced inverse effects at the spine, with comparable effects at the FN. The changes in FMI were associated with smaller, inverse changes at the spine in women and men, with a positive association between changes in FMI and hip BMD in postmenopausal women.

Table 4.

Regression Coefficients Between Changes in BMD (24 Months—Baseline) and Clinical Parameters and Biomarkers

	Men (n = 46–108)				Premenopausal Women (n = 28–29)				Postmenopausal Women (n = 55–100)				Sex × Predictor
	ΔSpine BMD (24–0)	P Value	ΔFN BMD (24–0)	P Value	ΔSpine BMD (24–0)	P Value	ΔFN BMD (24–0)	P Value	ΔSpine BMD (24–0)	P Value	ΔFN BMD (24–0)	P Value	Spine BMD	FN BMD
Age	−0.0001	.731	0.0003	.359	−0.0003	.825	−0.0008	.502	0.0004	.537	0.0003	.659	0.265	0.968
Baseline data
Weight, kg	0.0002	.547	−0.0005	.076	−0.00002	.976	0.0001	.908	0.0001	.697	0.0001	.766	0.462	0.292
BMI	0.0007	.503	−0.0015	.102	0.0003	.868	0.0014	.459	0.0009	.299	0.0008	.353	0.987	0.096
Waist circumference, cm	0.0001	.783	−0.0006	.065	−0.0004	.546	0.0002	.775	0.0004	.270	−0.0002	.645	0.974	0.580
Body fat, %	−0.0009	.229	−0.0001	.861	0.0001	.972	0.0002	.927	0.0005	.551	0.0001	.880	0.057	0.575
FMI	−0.0006	.671	−0.0011	.413	−0.0001	.972	0.0019	.527	−0.0005	.551	0.007	.560	0.164	0.253
LM, %	0.0009	.229	0.0001	.861	−0.0001	.972	−0.0002	.927	−0.0005	.551	−0.0001	.880	0.057	0.575
LMI	0.0033	.076	−0.0035	.043	0.0007	.844	0.0025	.486	0.0018	.357	0.0026	.205	0.198	0.097
Fat/LM ratio	−0.0405	.258	−0.0039	.907	−0.0006	.993	0.0105	.888	0.0161	.542	0.0086	.975	0.040	0.606
VATa	0.0016	.629	−0.0042	.111	N/Aa	N/Aa	N/Aa	N/Aa	0.0045	.131	−0.0015	.624	0.507	0.514
SATa	0.0026	.156	−0.0024	.099	N/Aa	N/Aa	N/Aa	N/Aa	−0.0013	.473	0.0003	.871	0.130	0.264
Physical activity	0.0304	.393	0.0303	.368	−0.0584	.463	0.1044	.189	−0.0165	.576	0.0399	.187	0.282	0.911
Follow-up data (24–0 mo)
ΔWeight	−0.0010	.006	0.0003	.322	−0.0001	.933	0.0032	.010	0.0016	.004	0.0022	<.0001	<0.0001	0.001
ΔBMI	−0.0032	.005	0.0011	.287	−0.0003	.941	0.0083	.011	0.0042	.004	0.0061	<.0001	<0.0001	0.003
ΔWaist circumference	−0.0008	.024	0.0004	.231	0.00005	.962	0.0017	.108	0.0009	.086	0.0015	.004	0.006	0.049
ΔBody fat, %	−0.0004	.026	0.0001	.444	−0.0001	.877	0.0013	.024	0.0007	.006	0.0011	<.0001	0.0004	<0.001
ΔFMI	−0.0037	.027	0.0014	.365	−0.0003	.952	0.0105	.015	0.0053	.007	0.0080	<.0001	0.0004	0.006
ΔLM, %	0.0004	.026	0.0006	.328	0.0001	.877	0.0039	.044	0.0007	.006	0.0029	<.001	<0.0001	0.025
ΔLMI	−0.0100	.001	0.003	.248	−0.0017	.869	0.0194	.0502	0.0119	.011	0.0149	.002	<0.0001	0.019
ΔFM/LM mass ratio	−0.0613	.171	0.0325	.444	−0.0012	.987	0.1909	.030	0.0996	.016	0.1619	<.0001	0.007	0.026
ΔVAT	−0.0060	.063	0.0017	.523	N/A*	N/A*	N/A*	N/A*	0.0120	.031	0.0230	<.0001	0.007	<0.001
ΔSAT	−0.0086	.003	0.00002	.995	N/A*	N/A*	N/A*	N/A*	0.0062	.037	0.0101	.0002	0.0003	0.005
ΔPhysical activity	0.0557	.152	0.0011	.976	0.0717	.414	−0.0835	.346	0.0672	.047	−0.0520	.139	0.863	0.310

Abbreviations: SAT, sc adipose tissue; N/A, not available. Menopausal status was self-reported at enrollment.

^aFor CT measures (VAT and SAT), n = 42 postmenopausal women, 0 premenopausal women, and 55 men.

Predictors of changes in BMD

In a post hoc analysis, we explored potential predictors of changes in BMD after weight loss, considering LM change (%), FM change (%), smoking status, physical activity change (Baecke Activity Factor), baseline calcium intake (for increments of 200 mg of dietary calcium), baseline total BMD, baseline age, change in dietary protein (%), baseline weight, baseline height, and use of contraceptives or hormone replacement therapy (in women only). In men, percentage change in BMD during the study was predicted by smoking status, LM change, and baseline calcium intake. The multivariable prediction equation (adjusted r² = 0.13) was: 0.93 [−2.81 (if current smoker) or −1.15 (if former smoker)] − 0.21 × LM change (%) + 0.21 × baseline calcium. In women, the percentage change in BMD was predicted by baseline BMI, smoking status, physical activity change, and LM change. The multivariable predication equation (adjusted r² = 0.13) was: −4.11 + 0.21 × BMI (baseline) [+4.25 (if current smoker) or −0.12 (if former smoker)] + 5.83 × physical activity change (Baecke Activity Factor) + 0.23 × LM change (%).

Discussion

In the POUNDS LOST study of four calorie-restricted diets, there were no differences in BMD changes among diets at 2 years. After weight loss, however, there were sex differences in changes in BMD. In postmenopausal women, there were small but significant decreases in spine and FN BMD, and in premenopausal women only a decrease in FN BMD was seen. In contrast, men showed an increase in spine BMD after weight loss, with no change at the FN. In men, the reduction in FM relative to LM was greater than in women. In postmenopausal women, reductions in LMI and FMI, measures normalized for height, were positively correlated with reductions in BMD at the spine and hip but inversely correlated with changes at the spine in men. In this large randomized controlled study, we demonstrated that changes in BMD after moderate weight loss may be sex-specific and that women lost BMD with weight loss whereas men gained BMD at the spine.

The elevated BMI in obese individuals has been associated with a high BMD compared with age-matched controls. Although the BMI is a composite measure of fat, muscle, and bone, in obese individuals the excess weight is generally attributed to more adipose tissue than lean muscle mass. In contrast, low BMI is a risk factor for low BMD and increased fracture risk in adults and in anorexic individuals through multiple mechanisms (21–23). Paradoxically, the skeletons of obese individuals relative to the loads on the skeleton are not stronger than nonobese individuals, and obesity is associated with an increased risk of fractures at the extremities (24, 25).

Most prospective trials assessing the effects of dietary interventions on bone loss in overweight and obese individuals are small and short-term (6 wk to 1 y), and most include only women or a small number of men (6, 26, 27). Although some studies report that weight loss of approximately 10% results in bone loss at the hip of 1–2% and 3–4% at sites enriched with trabecular bone (27, 28), the results vary according to menopausal status and sex. Hinton et al (27) showed in 40 overweight and obese premenopausal women that caloric restriction and aerobic exercise for 6 months (10% weight loss) resulted in spine and hip bone loss with an increase in bone turnover biomarkers. Of interest, these changes persisted after weight regain (27). In contrast, in another study among overweight premenopausal women on a weight loss program for 1 year, BMD Z-scores increased at the FN, TH, and spine (26). In our study, with a 6.4% weight loss, women lost BMD at both the spine and hip, whereas men exhibited an increase in BMD at the spine with no change at the hip. Although osteoarthritis may result in a factitious elevation in spinal BMD, the changes in spinal BMD were discordant in males and females enrolled in the POUNDS LOST study. Taken together, these results suggest a differential effect of weight loss on BMD at different sites in a manner dependent on both sex and menopausal status. With moderate weight loss, ie, 5% or more, data indicate a clinically significant increased risk of fractures in women and men (28–31). With extreme weight loss observed with bariatric procedures, there is also an increased risk of fractures, which may be related to the surgery, obesity, or weight loss (10). Thus, bone loss in women during weight loss interventions could adversely affect skeletal health and increase the risk of fractures. Because women are at a higher risk of osteoporosis than men, weight loss may adversely affect the skeletal health in females compared with men.

The amount of skeletal muscle and FM has important effects on BMD at different sites (32). In the CHAMP study, reduction in fat rather than LM closely correlated with decreases in BMD among men (33). In another large study of men and women, muscle mass was the strongest determinant of BMD and accounted for approximately 27% of the variance at the different skeletal sites. Physical activity per se was an additional independent predictor of bone mass, accounting for about 10% of the variance (34). In addition, in a cross-sectional study of women, both fat and lean body mass correlated with BMD, but the effects varied according to menopausal status, BMD site, and whether measurements were corrected for area or height (32). LM may indicate greater strength and more mechanical loading of bone than that conferred by adipose tissue. The lack of spinal bone loss in men and premenopausal women may thus relate to the greater muscle mass in men and younger women even in the presence of loss of muscle mass during calorie-restricted diets (35). In our prospective, randomized study, women lost BMD at the FN that was associated with LMI change, and men displayed a stable FN-BMD with a similar correlation with LMI. Furthermore, women showed a positive relationship between reductions in VAT and BMD at the spine and hip. Although the relatively small number of premenopausal women who adhered to diet for 2 years, the inverse relationship between changes in LMI and spine BMD in men and the greater loss of fat relative to LM in men as compared with postmenopausal women may suggest an effect of lean tissue on spinal BMD. Therefore, differences in the relationship between BMD and fat and lean tissue may vary according to age, sex, physical activity, and BMD site.

It is estimated that 40% of women and 20% of men aged 50 years or older will develop an osteoporotic fracture in their remaining lifetime (36). Bone loss in overweight and obese postmenopausal women on weight restrictive diets is, therefore, of concern especially in women who are at increased risk of osteoporosis. It has been suggested that calcium absorption is reduced with weight loss and that calcium supplementation may attenuate bone loss during weight loss in postmenopausal women (28). In contrast, no significant bone loss has been observed in premenopausal obese women after weight loss, even among those who consume a low-calcium diet (37).

Many other mechanisms may contribute to bone loss with weight loss in women. In addition to the effects of lean muscle mass and mechanical loading on bone, adipose tissue produces adipokines, cytokines, and estrogen, which may have differing effects on BMD (37, 38). There are complex interactions between changes in BMD, body composition, adipokines (eg, leptin, adiponectin), IGF-1 levels, bone remodeling, calcium homeostasis, and sex steroids (39). Further studies are necessary to determine whether sex differences in changes in BMD and body composition with weight loss may be accounted for by differences in bone turnover, mineral homeostasis, adipokines, or hormonal factors.

This study has several limitations. Although the POUNDS LOST trial incorporated rigorous behavioral approaches (1, 20), DXA measurements were available for only 78% and 56% of participants at 6 and 24 months, respectively. Nevertheless, these adherence rates exceed those of most long-term dietary intervention studies. To allow a meaningful amount of time to assess changes in BMD, outcomes were determined using data obtained from participants completing the 2-year interventions. In addition, we do not have measures of bone marrow adipose tissue that may potentially affect BMD. However, a recent study using magnetic resonance imaging found that changes in intraosseous soft tissue mass after weight loss had minimal effect on BMD (40). Furthermore, this study did not include measures of adipokines, bone turnover biomarkers, or calcium homeostasis, which requires further exploration. Strengths of this study include the use of several dietary approaches, containing four different concentrations of macronutrients, a large number of females and males, and repeated BMD measurements at both the spine and hip. To the best of our knowledge, no studies have included detailed measures of BMD and body composition measures during calorie-restricted, weight loss diets for 2 years.

In conclusion, long-term weight loss, regardless of the dietary macronutrient composition, results in complex, multifactorial effects on BMD. Sex-related factors, hormonal changes, differences in physical activity, as well as relative changes in fat and muscle mass may all contribute to the differential effects of weight loss on BMD in men and women at different ages. Additional studies are required to identify the ideal weight loss regimen combining efficient weight loss with positive effects on BMD and fracture risk.