Effect of Exercise Modality during Weight Loss on Bone Health in Older Adults with Obesity and Cardiovascular Disease or Metabolic Syndrome: A Randomized Controlled Trial

Abstract

The objective of this study was to determine the ability of either aerobic or resistance training to counter weight loss associated bone loss in older adults. 187 older adults (67 years, 70% women, 64% Caucasian) with obesity (body mass index: 34.5±3.7 kg/m²) and cardiovascular disease and/or metabolic syndrome were randomized to participate in an 18 month, community-based trial, with a follow-up assessment occurring at 30 months. Intervention arms included: weight loss alone (WL; 7–10% baseline weight), WL plus aerobic training (WL+AT), and WL plus resistance training (WL+RT). Dual energy x-ray absorptiometry (DXA) acquired total hip, femoral neck, and lumbar spine areal bone mineral density (aBMD) and trabecular bone score (TBS); and biomarkers of bone turnover (Procollagen Type 1 N-Terminal Propeptide, C-Terminal Telopeptide of Type 1 Collagen) were measured at baseline, six, 18, and 30 (aBMD and TBS, only) months. Computed tomography (CT) acquired hip and spine volumetric BMD (vBMD), cortical thickness, and bone strength was measured in a subset at baseline (n=55) and 18 months. Total hip aBMD was reduced by 2% in all groups at 18 months, with primary analysis showing no significant treatment effects for any DXA, biomarker, or CT outcome. After adjustment for weight loss and followed to 30 months, secondary analyses reveal total hip [−0.018 (−0.023 to −0.012) g/cm² vs −0.025 (−0.031 to −0.019) g/cm²; p=0.05] and femoral neck [−0.01 (−0.009 to 0.008) g/cm² vs −0.011 (−0.020 to −0.002) g/cm²; p=0.06] aBMD estimates were modestly attenuated in WL+RT compared to WL. Additionally, lumbar spine aBMD was increased in WL [0.015 (0.007 to 0.024) g/cm²] and WL+RT [0.009 (0.000 to 0.017) g/cm²] compared to WL+AT [−0.003 (−0.012 to 0.005) g/cm²]; both p≤0.01. Community-based exercise does not prevent bone loss during active weight loss in older adults; however, adding resistance training may help minimize long-term hip bone loss.

Trial Registration:

clinicaltrials.gov Identifier: NCT01547182.

INTRODUCTION

The prevalence of obesity and its detrimental health effects are increasing rapidly among older adults.⁽¹⁾ Medical complications associated with excess fat mass highlight the need to treat obesity in this age group;⁽²⁾ yet, weight loss recommendation remains controversial.^(3,4) Reluctance can be attributed, at least in part, to weight loss associated loss of bone mineral density (BMD), and potential exacerbation of age-related risk of osteoporotic fracture⁽⁵⁾-a leading cause of injury in older adults that significantly compromises both quality and expectancy of life. Indeed, observational data consistently link weight loss in late life with loss of BMD^(6–8) and increased fracture risk,^(7–11) regardless of weight loss intentionality or elevated baseline body weight.^(6,7) In agreement, meta-analytic data in middle-aged, overweight, and obese adults show that a single diet-induced weight loss intervention (of at least six months in duration) is associated with a small, but significant, decrease in total hip BMD (on the order of 0.010–0.015 g/cm²) and corresponding 10–15% increased risk of fracture.⁽¹²⁾ Emerging evidence now demonstrates that BMD is not fully recovered if weight is regained,^(13–16) and bone loss may even continue at an accelerated rate among those who are successful at long-term weight loss maintenance.⁽¹⁷⁾ These data provide impetus for the identification of weight loss strategies that minimize bone loss to yield maximal health benefit for older adults.

Several studies show a positive effect of exercise on BMD in weight-stable, older adults,^(18–20) and exercise may be an effective means to prevent bone loss during dietary weight loss. Some,^(21–25) but not all,^(26–30) studies support this hypothesis. Differing results may be due to varying exercise prescriptions, with recent clinical trial data suggestive of a superior ability of supervised resistance training to preserve bone mass with weight loss compared to aerobic training, at least in the short-term.^(31,32) Results warrant confirmation in long-term, community-based trials to inform real world weight management strategies. The current evidence base also consists mainly of studies conducted in younger populations [i.e., only four RCTs comparing WL alone to WL plus exercise have been conducted in adults aged >50 years^{(21,22,28,29)}] and using dual energy x-ray absorptiometry (DXA) acquired areal (a)BMD. Because exercise can improve bone quality (independent of aBMD),⁽³³⁾ and in lieu of adequately powered trials assessing fracture incidence, there is a need to comprehensively assess the effects of exercise modality during weight loss in older adults on multiple indices of bone health.

To address these research needs, we compare the effects of dietary induced weight loss alone (WL) to weight loss plus aerobic exercise training (WL+AT) or resistance exercise training (WL+RT) on DXA-derived bone health measures and circulating biomarkers of bone turnover in 187 older (60–79 years) adults with obesity undergoing an 18 month community-based weight loss intervention. This paper represents a secondary analysis of a previously reported study with dual primary outcomes of mobility and muscle strength.⁽³⁴⁾ Our primary hypothesis is that WL+RT will better attenuate BMD loss over 18 months than WL+AT or WL alone. In exploratory analyses, we control for the post-randomization effect of total weight loss, provide treatment effect estimates on computed tomography (CT) derived measures of bone density and quality, and report legacy effects observed the year following the conclusion of the intervention.

MATERIALS AND METHODS

Study Design and Participants

Details of the study design and methods are published.⁽³⁵⁾ Briefly, the Cooperative Lifestyle Intervention Program-II (CLIP II; NCT01547182) study was a multisite, single-blinded, parallel, randomized controlled trial (RCT) with an intervention that took place at YMCAs in Forsyth County, NC. The intervention was delivered and supervised by YMCA community partners with research staff serving as trainers/advisers in program delivery. A total of 249 (187 with baseline hip/spine aBMD; see Supplementary Figure 1) older adults with cardiovascular disease (CVD) or metabolic syndrome (MetS) and self-reported mobility disability were block randomized by wave via a computer-generated algorithm to one of three intervention groups: dietary induced weight loss alone (WL), WL plus aerobic exercise training (WL+AT), or WL plus resistance exercise training (WL+RT) for 18 months. After the intervention period, participants were transitioned to free living conditions and invited to return for a 30 month assessment visit. All participants provided written informed consent prior to study enrollment. The primary outcome paper, including detailed eligibility criteria and a CONSORT diagram, is published.⁽³⁴⁾

Intervention Descriptions

Weight Loss:

The three study arms received the same behaviorally-based weight loss intervention in phases: intensive (months 1–6), transition (months 7–12), and maintenance (months 13–18), with the goal of eliciting a 0.3 kg/week weight loss in the intensive phase (~330 kcal reduction/day) and a total weight loss of 7–10%. In accordance with the 2010 dietary guidelines,⁽³⁶⁾ the macronutrient breakdown of the diet was 20–25% protein, 25–30% fat, and 45–55% carbohydrate, and all participants were counseled to consume the recommended daily allowance of calcium and vitamin D.

Aerobic Exercise Training:

The AT intervention was an individually tailored, supervised walking program. The program frequency was four days/week, progressing to a duration goal of 45 minutes/day and walking intensity of 12–14 on the Borg Rating of Perceived Exertion (RPE) Scale.⁽³⁷⁾

Resistance Exercise Training:

The RT intervention was also individually tailored and involved a training frequency of four days/week, progressing to 45 minutes/day with an RPE of 15–18 as a target intensity for each resistance exercise. Participants completed three sets of 10–12 repetitions on eight upper and lower body machines with initial resistance determined from one repetition maximum (1RM) testing (goal of 75% of 1RM). Subsequently, when a participant completed 12 repetitions in the third set for two consecutive days, the resistance was increased to ensure progressive overload.

Measurements

Baseline Demographic and Intervention Process Information:

Participant age, sex, race/ethnicity, and medical history/co-morbid status were captured via self-report at the baseline visit. Height was assessed without shoes to the nearest 0.25 cm using a stadiometer (Health O Meter® Portrod) and body mass measured to the nearest 0.05 kg using a calibrated and certified digital scale (Health O Meter® Professional 349KLX). As previously described,⁽³⁴⁾ accelerometry data (collected over a period of seven continuous days at baseline and follow up visits) was used to assess change in minutes of moderate levels of physical activity, with intensity scaled from 1–9 (low to high) and a level greater than or equal to four defined as moderate to vigorous physical activity (MVPA). Time spent performing RT exercises was captured via self-report at six and 18 months.

DXA-acquired measures:

aBMD of the total hip, and femoral neck, and lumbar spine, and trabecular bone score (TBS) of the lumbar spine were determined by DXA (iDXA, GE Medical Systems, Madison, WI) at baseline (n=187), six (n=168), 18 (n=138), and 30 (n=108) months. All scans were performed in accordance with manufacturer recommended positioning and analyzed by an International Society of Clinical Densitometry certified DXA technologist blinded to intervention assignment. Coefficients of variation (CV) from repeated measurements (on the same individual by the same technician) at our institution are 1.21% for total hip aBMD, 1.82% for femoral neck aBMD, and 1.38% for lumbar spine aBMD.

Biomarkers of Bone Turnover:

Fasting blood samples were collected at baseline (n=184), six (n=166), and 18 (n=137) months via venipuncture following standard procedures with serum stored until analysis. In accordance with international recommendations,⁽³⁵⁾ bone formation marker Procollagen Type 1 N-Terminal Propeptide (P1NP), and bone resorption marker C-Terminal Telopeptide of Type 1 Collagen (CTX) were performed using commercially available ELISAs (NeoScientific, Cambridge, MA), as done previously.⁽³⁹⁾

CT-acquired measures:

CT scans of the femurs and lumbar spine were acquired on a 64-slice scanner (LightSpeed VCT, General Electric Medical Systems, Milwaukee, WI) on a subset of participants at baseline (n=55 of 187) and 18 month follow up (n=31). A 4-port bone mineral calibration phantom (Image Analysis, Columbia, KY) was imaged in every scan to allow for calculation of vBMD.⁽⁴⁰⁾ Processing of CT images for cortical thickness and finite element derived bone strength were performed according to published methodology.^(41,42) Briefly, the proximal femurs and L1-L4 vertebrae were segmented for each subject using thresholding, region growing, and manual editing (Mimics, Materialise, Plymouth, MI). Variable cortical thickness across the surface of the femurs and lumbar vertebrae was obtained using a validated algorithm (Stradwin v5.2, Cambridge University, UK).^(43–45) The algorithm fits a mathematical model constrained by a global cortical density and out-of-plane blur to Hounsfield Unit (HU) intensities measured from a line normal to the cortical surface that passes through the soft tissue, cortex, and trabecula. Point clouds of the inner and outer cortex surfaces were output, as well as cortical thicknesses (~14,000/femur; ~3,000/vertebrae). Thin-plate spline radial basis function interpolation with a relaxation algorithm was used to morph existing atlas FE models of the Global Human Body Models Consortium M50-O v4.4 femur and the Total Human Model for Safety AM50 v4.02 lumbar spine to each subject-specific geometry.^(42,46–49) Homologous landmarks collected using image segmentation and registration from analogous locations on the atlas and subject-specific geometries were used to derive an interpolation function and coefficients to morph the atlas FE models’ nodal coordinates to each subject geometry.⁽⁵⁰⁾ Subject-specific vBMD (ρ_CT) was used to adjust the elastic modulus of the femoral cortex (E, GPa; Eq. 1)^(51,52) and vertebral trabecular bone (E, MPa; Eq. 2)⁽⁵³⁾ in the FE models. A nearest neighbor mapping approach was applied to assign subject-specific cortical thicknesses to each cortical shell element node in the FE models.⁽⁴⁵⁾ Bone strength was estimated through simulation of the following experimental tests: single-limb stance, sideways fall, and quasi-static uniaxial vertebral compression.^(48,49) The peak fracture force or bone strength was defined as the peak force recorded between the impactor and femoral head or vertebral body, where fracture was modeled using element deletion to remove elements exceeding an effective plastic strain of 0.61% (vertebral trabecular bone)⁽⁵⁴⁾ or 0.88% (femoral cortex).⁽⁵⁵⁾

$$E=6.85{\left(\frac{0.877{\rho }_{CT,g/c{m}^3}+0.079}{0.6}\right)}^{1.49}$$ (1)

$$E=-34.7+3.23{\rho }_{CT,mg/c{m}^3}$$ (2)

Statistical Analyses

Descriptive statistics were calculated overall and by intervention group at baseline and by intervention and time at follow-up. Six and 18 month treatment effect estimates on weight, MVPA, time spent in RT, DXA-acquired outcome measures, and biomarkers of bone turnover were estimated (on all participants who had a baseline and a 6 or 18 follow up measurement) using a mixed model including group, time, and a group by time interaction, adjusted for sex, wave (i.e., recruitment phase), and baseline value of the outcome. Secondary analyses were conducted further adjusting for 18 month weight change and including of a gender*treatment interaction term in all model statements. 18 month treatment effect estimates on CT-acquired outcome measures were estimated using a general linear model, adjusted for sex, wave, and baseline value of the outcome, with and without weight change over 18 months. Overall legacy effects on DXA-acquired outcome measures, including the 30 month time point, were estimated using a mixed model including group, time, and a group by time interaction, adjusted for sex, wave, and baseline value of the outcome, with and without weight change over 30 months.

RESULTS

Baseline Participant Characteristics and Intervention Compliance

Participant enrollment occurred from July 2012 through March 2014, with active follow up through September 2016. Baseline descriptive characteristics, detailed by group and overall, are presented in Table 1. Average age was 66.9±4.8 years, 70% were women, 32% were African American, and over half (56%) reported post-secondary education. Average BMI was 34.5±3.7 kg/m², with 41% presenting with osteopenia (39%) or osteoporosis (2%), based on World Health Organization classification guidelines.⁽⁵⁶⁾ Baseline descriptive characteristics were similar for CT subgroup participants (n=55; see Supplementary Table 1).

Table 1.

Baseline descriptive characteristics according to treatment group.

	WL (N=60)	WL+AT (N=67)	WL+RT (N=60)	Overall (N=187)
Age (years)	66.3 ± 4.6	67.3 ± 5.1	67.0 ± 4.5	66.9 ± 4.8
Female, n (%)	44 (72.1)	48 (71.6)	40 (66.7)	132 (70.2)
Race/Ethnicity, n (%)
African American	23 (37.7)	25 (37.3)	13 (21.7)	61 (32.4)
Hispanic	1 (1.6)	0 (0.0)	1 (1.7)	2 (1.1)
White	35 (57.4)	42 (62.7)	44 (73.3)	121 (64.4)
Other/Mixed/Missing	2 (3.3)	0 (0.0)	2 (3.3)	4 (2.1)
Highest Level of Education, n (%)
Less than high school diploma	1 (1.6)	3 (4.5)	1 (1.7)	5 (2.7)
High school/some college	23 (37.7)	24 (35.8)	31 (51.7)	78 (41.5)
Associate’s degree or higher	37 (60.7)	40 (59.7)	28 (46.7)	105 (55.9)
Comorbidities, n (%)
CVD History	40 (65.6)	49 (73.1)	42 (70.0)	131 (69.7)
Diabetes	10 (16.4)	16 (23.9)	13 (21.7)	39 (20.7)
Arthritis	33 (55.0)	33 (49.3)	38 (66.7)	104 (56.5)
Hypertension	41 (67.2)	54 (80.6)	44 (73.3)	139 (73.9)
Any Cancer	18 (29.5)	7 (10.4)	13 (22.0)	38 (20.3)
Metabolic Syndrome	40 (65.6)	37 (55.2)	37 (61.7)	114 (60.6)
BMI (kg/m2)	35.0 ± 4.1	33.9 ± 3.5	34.6 ± 3.5	34.5 ± 3.7
Total Body Mass (kg)	97.2 ± 18.2	93.8 ± 13.8	97.5 ± 14.8	96.1 ± 15.7
DXA-derived Osteoporosis, n (%)	3 (4.9)	0 (0.0)	0 (0.0)	3 (1.6)
DXA-derived Osteopenia, n (%)	22 (36.1)	24 (35.8)	27 (45.0)	73 (38.8)
Moderate-Vigorous Physical Activity (min/wk)	82.6 ± 64.4	118.3 ± 141.7	93.3 ± 102.1	98.6 ± 109.0

Continuous data are presented as means±SD and categorical data are presented as n (%).

In accordance with the goals of the trial, all three treatment groups lost significant weight; yet, both weight loss plus exercise groups experienced significantly and similarly augmented total weight loss over the 18 month period as compared to weight loss alone [mean (95% CI): WL+RT: −10.1 (−12.0 to −8.2) kg and WL+AT: −9.9 (−11.8 to −7.9) kg versus WL: −5.7 (−7.9 to −3.8) kg; both p<0.01]. Over the course of the intervention, average accelerometer-based MVPA significantly increased from baseline by 218 (180 to 255) minutes/week in the WL+AT group [versus −8 (−47 to 31) minutes/week and 22 (−14 to 58) minutes/week in the WL and WL+RT groups, respectively], and self-reported time spent performing resistance training significantly increased from baseline by 182 (170 to 194) minutes/week in the WL+RT group [versus 3 (−9 to 15) minutes/week and 2 (−10 to 14) minutes/week in WL and WL+AT, respectively]. Additionally, and as reported in the main outcome paper, WL plus exercise groups improved 400-m walk time (~17 second reduction) and relative knee extensor strength (~15% improvement) as compared to WL alone, with no difference between the AT or RT groups.⁽³⁴⁾

Immediate Intervention Effect on DXA, Biomarker, and CT Metrics of Bone Health

Unadjusted baseline means and standard deviations, as well as six and 18 month model adjusted intervention effect estimates on change in DXA-derived bone health measures are presented in Table 2. Overall, no significant intervention effects were observed on any DXA-acquired measure; however, total hip BMD was reduced by roughly 2% in all groups at 18 months [WL: −0.023 (−0.033 to −0.014) g/cm²; WL+AT: −0.027 (−0.036 to −0.018) g/cm²; WL+RT: −0.025 (−0.034 to −0.016) g/cm²]. Similarly, biomarkers of bone turnover were unchanged over time or by group (data not shown). Further adjustment for differential group weight loss did not materially alter DXA or biomarker results, and no significant gender*treatment interaction was observed for any outcome.

Table 2.

Intervention effects on change in DXA-derived bone health measures.

	WL (n: 53–60)	WL+AT (n: 58–67)	WL+RT (n: 57–60)	p-value
Total Hip aBMD (g/cm2)
Baseline	1.009±0.121	1.016±0.133	1.010±0.146
Change at 6 months	−0.015 (−0.023 to −0.006)	−0.013 (−0.020 to −0.005)	−0.015 (−0.024 to −0.007)	0.88
Change at 18 months	−0.023 (−0.033 to −0.014)	−0.027 (−0.036 to −0.018)	−0.025 (−0.034 to −0.016)	0.84
Femoral Neck aBMD (g/cm2)
Baseline	0.953±0.119	0.951±0.116	0.951±0.127
Change at 6 months	−0.004 (−0.017 to 0.009)	0.000 (−0.012 to 0.012)	0.004 (−0.009 to 0.016)	0.71
Change at 18 months	−0.009 (−0.023 to 0.005)	0.004 (−0.010 to 0.017)	0.004 (−0.009 to 0.017)	0.30
Lumbar Spine aBMD (g/cm2)
Baseline	1.263±0.186	1.293±0.220	1.279±0.198
Change at 6 months	0.012 (0.001 to 0.023)	0.003 (−0.007 to 0.013)	0.009 (−0.001 to 0.019)	0.41
Change at 18 months	0.012 (0.001 to 0.024)	−0.006 (−0.017 to 0.006)	0.008 (−0.003 to 0.019)	0.06
Trabecular Bone Score
Baseline	1.393±0.111	1.421±0.097	1.425±0.116
Change at 6 months	0.041 (0.017 to 0.066)	0.029 (0.006 to 0.052)	0.037 (0.014 to 0.059)	0.70
Change at 18 months	−0.003 (−0.028 to 0.023)	−0.007 (−0.032 to 0.018)	−0.003 (−0.028 to 0.022)	0.95

Unadjusted baseline means ± SD. Six and 18 month change estimates (95% CI) were estimated using a mixed model including group, time, and a group by time interaction and adjusted for sex, wave, and baseline value of the outcome. Sample size ranges by group and time are as follows: 6-month n=43–50, 53–62, and 53–56; 18-month n=39–43, 40–47, and 43–48 for WL, WL+AT, and WL+RT, respectively.

Exploratory 18 month intervention effects on the subset of participants assessed for CT-derived metrics of bone density and quality are presented in Table 3. Like aBMD at the hip, vBMD was significantly reduced in all groups; however, this decline tended to be less in WL+RT participants [−0.015 (−0.024 to −0.006) g/cm³] compared to WL [−0.027 (−0.036 to −0.019) g/cm³] or WL+AT [−0.029 (−0.037 to −0.020) g/cm³] participants. Similar trends were observed for change in vBMD at the femoral neck and cortical thickness estimates at all sites, with greatest losses noted in the WL group, followed by the WL+AT group, and then the WL+RT group. As with aBMD data, treatment effect estimates of lumbar spine vBMD indicate a trend toward reduced density in the WL+AT group compared WL or WL+RT groups [WL+AT: −0.014 (−0.027 to −0.001) g/cm³ versus WL: −0.005 (−0.022 to 0.012) g/cm³ and WL+RT: −0.004 (−0.019 to 0.011) g/cm³]. As with DXA and biomarker data, further adjustment for differential group weight loss over 18 months did not materially alter CT results (data not shown).

Table 3.

Intervention effects on change in CT-derived bone density and quality measures.

	WL (n=15–17)	WL+AT (n=17–19)	WL+RT (n=16–19)	p-value
Total Hip vBMD (g/cm3)
Baseline	0.297±0.032	0.314±0.047	0.292±0.023
Change at 18 months	−0.027 (−0.036 to −0.019)	−0.029 (−0.037 to −0.020)	−0.015 (−0.024 to −0.006)	0.06
Femoral Neck vBMD (g/cm3)
Baseline	0.308±0.035	0.326±0.059	0.305±0.029
Change at 18 months	−0.019 (−0.028 to −0.009)	−0.011 (−0.020 to −0.001)	−0.006 (−0.017 to 0.004)	0.22
Lumbar Spine vBMD (g/cm3)
Baseline	0.126±0.032	0.133±0.058	0.120±0.020
Change at 18 months	−0.005 (−0.022 to 0.012)	−0.014 (−0.027 to −0.001)	−0.004 (−0.019 to 0.011)	0.54
Total Hip Cortical Thickness (mm)
Baseline	2.002±0.248	2.086±0.213	2.000±0.255
Change at 18 months	−0.049 (−0.075 to −0.023)	−0.042 (−0.067 to −0.017)	−0.040 (−0.067 to −0.014)	0.87
Femoral Neck Cortical Thickness (mm)
Baseline	1.842±0.215	1.937±0.252	1.830±0.253
Change at 18 months	−0.031 (−0.061 to −0.000)	−0.028 (−0.058 to 0.001)	−0.020 (−0.051 to 0.011)	0.88
Lumbar Spine Cortical Thickness (mm)
Baseline	1.117±0.162	1.175±0.139	1.178±0.158
Change at 18 months	−0.030 (−0.116 to 0.056)	0.005 (−0.058 to 0.067)	0.010 (−0.061 to 0.081)	0.74
Femoral Stance Strength (kN)
Baseline	6.006±0.971	6.164±1.147	6.362±0.890
Change at 18 months	−0.168 (−0.265 to −0.072)	−0.146 (−0.237 to −0.054)	−0.141 (−0.239 to −0.044)	0.91
Femoral Fall Strength (kN)
Baseline	2.051±0.370	2.037±0.491	2.121±0.298
Change at 18 months	−0.059 (−0.109 to −0.009)	−0.030 (−0.079 to 0.018)	−0.044 (−0.096 to 0.008)	0.70
Vertebral Compression Strength (kN)
Baseline	3.686±1.006	3.920±1.307	3.449±0.781
Change at 18 months	−0.268 (−0.621 to 0.084)	−0.358 (−0.622 to −0.095)	−0.051 (−0.349 to 0.248)	0.30

Unadjusted baseline means ± SD. 18 month change estimates (95% CI) adjusted for sex, wave, and baseline value of the outcome. Follow up sample size ranges by group are n=7–8, 9–13, and 8–10 for WL, WL+AT, and WL+RT, respectively.

Intervention Legacy Effects on Weight and DXA Metrics of Bone Health

Data from the 30 month assessment visit show a diminished but persistent treatment effect on overall body weight, with the weight loss plus exercise groups maintaining greater weight loss than weight loss alone [WL+AT: −7.1 (−9.4 to −4.8) kg and WL+RT: −6.6 (−9.0 to −4.3) kg versus WL: −3.4 (−5.7 to −1.5) kg; both p<0.01]. In secondary analyses, we present intervention legacy effects on the DXA-acquired bone metrics, before and after adjustment for the post-randomization effect of change in total weight. As presented in Figure 1, despite significant weight regain in all groups from 18 to 30 months (p<0.01), most aBMD and TBS estimates continued to decline. The exception was lumbar spine BMD in the WL-only group, which was increased above baseline at all follow up time points.

Figure 1.

Percent change (95% CI) in DXA-acquired (A) Total Hip aBMD, (B) Femoral Neck aBMD, (C) Lumbar Spine aBMD, and (D) Trabecular Bone Score by group over 30 months of follow up. Model estimates were generated using a mixed model including group, time, and a group by time interaction and adjusted for sex, wave, and baseline value of the outcome.

Overall treatment effects on DXA-acquired bone metrics, including 30 month data, are presented in Table 4. As with 18 month data, reduction in total hip BMD was observed in all groups over time; however, after adjustment for change in total weight (Model 2), a modest treatment effect was observed between WL and WL+RT groups [WL: −0.025 (−0.031 to −0.019) g/cm² versus WL+RT: −0.018 (−0.023 to −0.012) g/cm²; p=0.05]. A trend was also observed for femoral neck aBMD when contrasting WL and WL+RT group estimates [WL: −0.011 (−0.020 to −0.002) g/cm² versus WL+RT: −0.000 (−0.009 to 0.008) g/cm²; p=0.06]. Lastly, an overall treatment effect persisted for lumbar spine aBMD with significant increases observed in WL and WL+RT groups, compared to WL+AT, with and without adjustment for overall weight change (both p≤0.01).

Table 4.

Overall intervention effects on DXA-derived bone health measures, including 30 month legacy data.

	WL	WL+AT	WL+RT	p-value
				WL+AT vs. WL-Only	WL+RT vs. WL-Only	WL+AT vs. WL+RT
Total Hip aBMD (g/cm2)
Model 1	−0.020 (−0.026 to −0.014)	−0.023 (−0.029 to −0.018)	−0.020 (−0.026 to −0.014)	0.45	0.98	0.46
Model 2	−0.025 (−0.031 to −0.019)	−0.022 (−0.028 to −0.017)	−0.018 (−0.023 to −0.012)	0.39	0.05	0.24
Femoral Neck aBMD (g/cm2)
Model 1	−0.010 (−0.018t to −0.001)	−0.003 (−0.011 to 0.005)	−0.001 (−0.009 to 0.007)	0.21	0.13	0.76
Model 2	−0.011 (−0.020 to −0.002)	−0.002 (−0.011 to 0.006)	−0.000 (−0.009 to 0.008)	0.12	0.06	0.71
Lumbar Spine aBMD (g/cm2)
Model 1	0.016 (0.007 to 0.024)	−0.003 (−0.012 to 0.005)	0.009 (0.000 to 0.017)	<0.01	0.14	0.01
Model 2	0.015 (0.007 to 0.024)	−0.003 (−0.012 to 0.005)	0.009 (0.000 to 0.017)	<0.01	0.16	0.01
Trabecular Bone Score
Model 1	0.010 (−0.007 to 0.028)	0.006 (−0.012 to 0.023)	0.002 (−0.015 to 0.019)	0.63	0.38	0.70
Model 2	0.011 (−0.007 to 0.029)	0.005 (−0.012 to 0.023)	0.002 (−0.015 to 0.019)	0.55	0.33	0.70

Model 1 estimates (95% CI) were generated using a mixed model including group, time, and a group by time interaction and adjusted for sex, wave, and baseline value of the outcome. Model 2 estimates further adjust for weight change. 30-month sample size ranges by group are n=33–35, 31–38, and 33–35 for WL WL+AT, and WL+RT, respectively.

DISCUSSION

The objective of the current investigation was to determine the effect of exercise modality during dietary induced weight loss on multiple indices of bone health in older adults with obesity. As reported in the main outcome paper,⁽³⁴⁾ total weight loss was greatest when exercise was coupled to dietary induced weight loss (~9% versus 5%). Here we primarily report a significant 2% reduction in total hip aBMD in all groups over the 18 month intervention period, with no measurable treatment effect on any DXA-acquired or biomarker outcome measure. However, after consideration of total weight lost and when followed out to 30 months, a modest, but consistent, trend from secondary analyses suggests that resistance training may attenuate long term weight loss associated bone loss at the hip region. Additionally, weight loss alone or with resistance, but not aerobic, training was found to increase lumbar spine aBMD by 30 months of follow up. Preliminary estimates of CT-acquired measures of volumetric density and cortical thickness are also suggestive of a potential osteoprotective effect of resistance training during weight loss, although findings warrant confirmation in an appropriately powered trial. Lastly, we report that for several DXA-acquired outcomes, bone loss continued despite weight regain in the 12 months following cessation of the interventions. Although an emerging area, our findings are in general agreement with other studies monitoring legacy effects of weight loss on bone health, and suggest that declining bone mass may persist after weight loss ceases, or reverses.¹³⁻¹⁶

Large prospective studies consistently show that weight loss is associated with higher hip fracture risk in older adults,⁽⁵⁷⁾ even among those who are overweight or obese.⁽⁷⁾ Importantly, newly published results from the Look AHEAD study (the only RCT, to our knowledge, which has examined the effect of long term intentional weight loss on fracture incidence in overweight and obese adults) confirms that 6–9% weight loss achieved and maintained over nearly a decade increases frailty fracture by 39%,⁽⁵⁸⁾ with concomitant loss in BMD^(59,60)-although it is worth noting that overall fracture incidence was the same for those randomized to weight loss versus education control. While it remains to be determined if or how strongly clinicians should counsel older patients with obesity about the impact of intentional weight loss on skeletal health,⁽⁵⁾ results from this study, and others,^(31,32,61) point toward a recommendation to couple resistance training with dietary induced weight loss to optimize the musculoskeletal response. Of note, a recently published six month RCT in older adults with obesity observed that despite similar weight loss (−9%), loss of BMD at the total hip was −2.7% and −1.1% in participants randomized to weight loss plus aerobic training and weight loss plus the combination of aerobic and resistance training, respectively, yet remained stable (−0.6%) in participants randomized to weight loss plus resistance training.⁽³²⁾ Clinically, this suggests that RT alone may be even more beneficial to bone health than the combination of RT and AT; although certainly, final exercise recommendation must be considered in light of individual treatment goals.

Novel findings presented here suggest the benefit of progressive RT may extend to central measures of volumetric bone density and quality, although estimates were primarily generated for the purpose of sufficiently powering subsequent trials and should not be considered definitive. Future studies should include sophisticated (i.e., CT) metrics of bone density and quality to elucidate independent and combined effects of weight loss and exercise on bone health, as they may independently impact skeletal metabolism. For instance, close inspection of changes in total hip aBMD from a RCT where older adults with obesity were randomized to weight stable, weight loss, exercise, or the combination of weight loss and exercise conditions,⁽²²⁾ does not support an interaction between exercise and weight loss arms; that is, the main effect of weight loss on bone (weight loss versus control: ~2.5% loss) was similar in magnitude and in opposition to the main effect of exercise on bone (weight loss plus exercise versus exercise: ~2.5% gain). This may indicate that exercise does not “prevent” bone loss induced by weight loss per se, rather that weight loss and exercise are affecting (perhaps trabecular and cortical) bone differently, which would not be evident from aBMD alone.

Confirmatory findings presented here, showing roughly 1–2% bone loss (particularly at the hip) with 10% weight loss,⁽¹²⁾ even among WL+RT participants, and failed recovery of bone mass with weight regain^(13–16) suggest that skeletal remodeling is affected by factors other than change in loading forces. Such mechanisms may include, but are not limited to, nutritional deficiencies and neuro-hormonal alterations,⁽⁶⁾ and is supported by data showing a differential response in BMD to caloric restriction induced versus exercise induced weight loss.⁽⁶²⁾ In order to design effective bone sparing intervention strategies, better understanding of the biology of weight loss-associated bone loss is needed (particularly in older adults, for whom the absolute risk of fracture and catabolic effects of weight loss on bone loss are greatest). Additionally, if minimizing bone loss during active weight loss proves necessary to offset long-term skeletal fragility, then our results suggest that resistance exercise alone – at least at the level achieved in our community based setting-may not provide sufficient anabolic stimulus to fully prevent the catabolic effects of weight loss on bone, and may need to be coupled with other intervention strategies to maximize skeletal benefit.

Strengths of this study include the RCT design, long term follow up, and comprehensive assessment of changes in bone health. As previously stated, as secondary endpoints, the CLIP II trial was not powered to detect differences in bone outcome measures. Additionally, our exercise intervention was not specifically designed to load bone using a high-intensity progressive resistance training and/or impact loading protocol⁽⁶³⁾ and, due to the community-based nature of the trial, we did not capture direct measures of mechanical stress to quantify exercise exposure. Although future research aimed at optimizing the skeletal response to weight loss may adapt slightly different exercise prescriptions, this protocol was designed to be delivered in a community-based setting and is effective in preserving muscle mass,⁽⁶¹⁾ which is the primary source of anabolic mechanical stimuli for bone tissue. We acknowledge that adjustment for weight change is a post-randomization effect, and are careful to present trial findings (utilizing the same analytic strategy as the main outcome paper), first. However, we were primarily concerned as to whether exercise could mitigate weight loss associated bone loss which necessitated standardization of weight change across groups. Although DXA remains the gold standard for clinically assessing bone health, it is susceptible to obesity and weight loss induced measurement error, and is limited in its ability to discern between true increases in BMD versus degenerative change. Indeed, discordant findings between our spine aBMD and vBMD results may be due to radiographic artifact in our sample. Lastly, the lack of effect for CTX and P1NP may be due to a mismatch in time course maximal expression of these biomarkers and blood collection visits in this study, as meta-analytic data suggest peak expression of these biomarkers occurs 2–3 months post intervention.⁽¹²⁾ And, although selected based on international consensus,⁽³⁸⁾ measurement of P1NP and CTX provide only a small glimpse into the mechanistic underpinnings of weight loss and exercise on bone health. Future studies may consider expanding this biomarker panel to include novel indicators of bone health, such as sclerostin, which is increased in the context of weight loss but may be negated by concurrent exercise training.⁽⁶⁴⁾

In sum, we report that community-based resistance training or aerobic training was unable to prevent bone loss during active weight loss in older adults; however, adding resistance training may help to minimize long-term bone loss, particularly at the hip. Future research efforts should aim to uncover the underlying mechanisms of weight loss induced bone loss, so that safe and effective strategies can be designed to preserve all aspects of bone health in this population.