Bone Density and Trabecular Morphology at Least 10 Years After Gastric Bypass and Gastric Banding

ABSTRACT

Roux‐en‐Y gastric bypass (RYGB) instigates high‐turnover bone loss in the initial 5 years after surgery, whereas skeletal changes after adjustable gastric banding (AGB) are less pronounced. Long‐term skeletal data are scarce, and the mechanisms of bone loss remain unclear. We sought to examine bone density and microarchitecture in RYGB and AGB patients a decade after surgery and to determine whether prior published reports of bone loss represent an appropriate adaptation to new postsurgical weight. In this cross‐sectional study, 25 RYGB and 25 AGB subjects who had bariatric surgery ≥10 years ago were matched 1:1 with nonsurgical controls for age, sex, and current body mass index (BMI). We obtained bone mineral density (BMD) by dual‐energy X‐ray absorptiometry (DXA), volumetric BMD and microarchitecture by high‐resolution peripheral quantitative computed tomography (HR‐pQCT), trabecular morphology by individual trabecular segmentation, and metabolic bone laboratory results. As compared with BMI‐matched controls, RYGB subjects had significantly lower hip BMD, and lower total volumetric BMD at the distal radius and tibia. Substantial deficits in cortical and trabecular microarchitecture were observed in the RYGB group compared to controls, with reduced trabecular plate bone volume fraction and estimated failure load at both the radius and tibia, respectively. Bone turnover markers CTX and P1NP were 99% and 77% higher in the RYGB group than controls, respectively, with no differences in serum calcium, 25‐hydroxyvitamin D, or parathyroid hormone. In contrast, the AGB group did not differ from their BMI‐matched controls in any measured bone density, microarchitecture, or laboratory parameter. Thus, RYGB, but not AGB, is associated with lower than expected hip and peripheral BMD for the new weight setpoint, as well as deleterious changes in bone microarchitecture. These findings suggest that pathophysiologic processes other than mechanical unloading or secondary hyperparathyroidism contribute to bone loss after RYGB, and have important clinical implications for the long‐term care of RYGB patients. © 2020 American Society for Bone and Mineral Research.

BARIATRIC SURGERY, BONE DENSITY, BONE MICROARCHITECTURE, INDIVIDUAL TRABECULAR SEGMENTATION, SKELETAL UNLOADING

Introduction

Bone loss is a known consequence of bariatric surgery procedures such as Roux‐en‐Y gastric bypass (RYGB).^(1–3) Studies have shown the rapid development of high‐turnover bone resorption in the initial years after RYGB, with bone resorption markers remaining elevated throughout at least 5 years after surgery.^(4–8) Declines in areal and volumetric bone mineral density (BMD) of approximately 10% at axial sites and 15% at peripheral sites have been captured by multiple imaging modalities in longitudinal studies.^(7–13) However, studies also indicate that different types of bariatric procedures may have different skeletal effects. In particular, short‐term studies of the purely restrictive adjustable gastric banding (AGB) procedure have found modest effects, if any, on bone turnover^(14–16) and BMD.^(15–17) No studies to date have explored whether bone microarchitecture, including trabecular morphology, might detect more subtle alterations after AGB. Furthermore, most studies have focused on changes within the first 2 years after bariatric surgery and little is known about the long‐term effects of any bariatric surgery procedure on bone health. Given that most patients receive bariatric surgery during midlife (median age 46 years),⁽¹⁸⁾ it is critical to understand the lifelong skeletal implications of these procedures.

In addition, it is unknown whether the accelerated bone loss observed after bariatric surgery is due to physiologically‐appropriate adaptations to surgery‐induced weight loss, or whether it represents a pathologic condition.^(2,3,19) Mechanical unloading of the skeleton occurs during all forms of weight loss,⁽²⁰⁾ and has been proposed as an explanation for the high‐turnover bone loss observed after bariatric surgery. Thus, it is important to clarify whether surgery‐induced bone loss is weight‐appropriate. This information will provide insight into the mechanisms of bone loss and have implications for the clinical care of bariatric patients.

We therefore performed a cross‐sectional study of BMD and bone microarchitecture, including trabecular morphology, in adults who received RYGB or AGB surgery at least 10 years ago compared to nonsurgical controls matched for age, sex, and current postoperative body mass index (BMI). The goal was to compare bone parameters after different bariatric procedures with matched controls to investigate whether bariatric patients achieved long‐term physiologic expectations for postoperative weight. We hypothesized that subjects who previously received RYGB, but not AGB, would have lower bone density and worse indices of skeletal microarchitecture and trabecular morphology compared to their BMI‐matched controls.

Patients and Methods

Study subjects

We recruited 25 adults (age ≥ 21 years) who had received Roux‐en‐Y gastric bypass (RYGB) surgery and 25 adults who had received adjustable gastric banding (AGB) surgery ≥10 years ago from two US academic medical centers. Each surgical subject was matched 1:1 with a nonsurgical control subject for age ± 5 years, sex, and current BMI ± 3 kg/m². In addition, we attempted to recruit controls of similar menopause status (women only), and race/ethnicity. Healthy controls were recruited through our clinical Weight Center as well as an established institutional recruitment website. Bariatric surgery subjects were excluded if they subsequently underwent a revision to another type of bariatric surgery, or in the case of AGB subjects, if the band was removed within 5 years of the initial surgery. Additional exclusion criteria were disorders known to affect bone metabolism (with the exception of osteopenia/osteoporosis), use of bone‐modifying medications (eg, glucocorticoids in the previous 3 months, osteoporosis or other antiresorptive medications in the previous year), previous cancer treated with chemotherapy or radiation, and early menopause. Race/ethnicity and medical history was provided by self‐report. Menopause status was determined as lack of menstrual periods in the previous 12 months, or follicle‐stimulating hormone (FSH) elevation greater than two times normal if the subject had a hysterectomy. History of diabetes (ever) was defined by self‐report and review of current medications. Fracture history was ascertained as self‐reported fractures occurring after age 20 years, excluding fractures of the face/skull, fingers, and toes. The study was approved by the Partners Human Research Committee, and all subjects provided written informed consent.

Dual‐energy X‐ray absorptiometry

Areal bone mineral density (aBMD; g/cm²) was measured by dual‐energy X‐ray absorptiometry (DXA) at the posterior–anterior (PA) lumbar spine (L₁–L₄), total hip, femoral neck, 1/3 radius, and subtotal body (Hologic Discovery A; Hologic, Inc., Marlborough, MA, USA). Subtotal body composition was also obtained, including measurements of lean mass (kg) and fat mass (kg). Independent experts reviewed DXA scans for quality control. We performed a short‐term reproducibility study at our bone density center in which 30 healthy volunteers were scanned twice with repositioning between scans. in vivo scanning precision at our institution was determined to be 0.005, 0.006, and 0.009 g/cm² for PA spine, total hip, and femoral neck, respectively.

High‐resolution peripheral quantitative CT

Volumetric bone mineral density (vBMD; g/cm³) and bone microarchitecture were assessed at the distal radius and tibia using high‐resolution peripheral quantitative CT (HR‐pQCT) (XtremeCT; SCANCO Medical AG, Brüttisellen, Switzerland). A standard region of interest (ROI) was used to determine the scan site and placed a fixed distance 9.5 and 22.5 mm from the radial and tibial endplates, respectively. Cortical and trabecular density and microarchitecture were calculated using standard analysis software (SCANCO software version V6.0). A semiautomated cortical bone segmentation technique was used to characterize cortical microarchitecture in greater detail.⁽²¹⁾ Linear micro‐finite element analysis (μFEA) was used to calculate estimated failure load, a measure of bone strength, in response to simulated uniaxial compression.⁽²²⁾ Only scans with minimal motion artifact (1 to 2 on a 3‐point scale) were included in analyses.

Individual trabecular segmentation

Individual trabecular segmentation (ITS) provided more detailed analyses of trabecular bone morphology. This technique allows for distinction between plate‐like and rod‐like trabecular structures, as well as the number, thickness, and length of these components.^(23,24) The orientation of the trabecular network is described using axial bone volume fraction. Better bone strength is generally associated with a higher proportion of trabecular plate‐like morphology, as well as larger bone volume in axial alignment. At our center, the reproducibility (as expressed by root‐mean square coefficient of variation) of ITS variables ranges from 0.9% to 7.5% at the radius, and 1.9% to 9.3% at the tibia.

Biochemical measurements

Fasting morning serum was collected for laboratory evaluations. Serum collagen type I cross‐linked C‐telopeptide (CTX), a measurement of bone resorption, was assessed by immunoradiometric assay (Immunodiagnostic Systems, Fountain Hills, AZ, USA) with intraassay variation of 5.2% to 6.8% and interassay variation of 5.6% to 7.4%. Procollagen type I N‐terminal propeptide (P1NP), a measure of bone formation, was assessed by radioimmunoassay with an intraassay variation of 3.5% to 5.3% and interassay variation of 3.6% to 5.4% (Orion Diagnostics, Espoo, Finland). Serum calcium (Quest Diagnostics, Secaucus, NJ, USA), 25‐hydroxyvitamin D (liquid chromatography–tandem mass spectrometry [LC‐MS/MS]; intraassay variation of <5% relative standard deviation [RSD] and interassay variation of <8% RSD), parathyroid hormone (PTH) (Access chemiluminescent immunoassay; Beckman Coulter, Fullerton, CA, USA; intraassay variation of 1.6% to 2.6% and interassay variation of 2.8% to 5.8%; normal range 12 to 88 pg/mL), and fasting insulin (Access chemiluminescent immunoassay; Beckman Coulter, Fullerton, CA, USA; intraassay variation 2.0% to 4.2% and interassay variation of 3.1% to 5.6%) were measured. Homeostatic Model Assessment of Insulin Resistance (HOMA‐IR) was calculated as fasting serum glucose (mg/dL) multiplied by insulin (mIU/mL) divided by 405, as a measure of insulin resistance.

Bionutrition measurements

Measurements of height and weight were obtained using a wall‐mounted Harpenden stadiometer (Seritex, Inc, East Rutherford, NJ, USA) and digital scale (Tanita BWB‐800; Tanita Corporation of America, Inc, Arlington Heights, IL, USA), respectively. Weight loss was calculated as the difference between measured weight on the day of study visit and either the preoperative weight from the surgical note (available in 40% of RYGB and 52% of AGB subjects) or self‐reported preoperative weight for the remaining subjects. Postoperative nadir weight was also assessed using self‐report. A registered dietitian administered the Calcium and Vitamin D Food Frequency Questionnaire to determine total daily calcium and vitamin D intake, encompassing both dietary intake and supplements.⁽²⁵⁾ Leisure and occupational physical activity were measured using the Modified Activity Questionnaire.⁽²⁶⁾

Statistical analysis

Data are reported as mean ± SD unless otherwise noted. The RYGB and AGB groups were analyzed separately in comparison with their respectively matched control groups. Differences in bone density and bone microarchitectural outcomes between each surgical group and their respective controls were compared using paired t tests and/or Fisher's exact tests. Pearson's correlations assessed univariate relationships between bone outcomes to weight loss and laboratory parameters to evaluate possible mediators of bone loss in the surgical group. Multivariate linear regression modeling was performed to adjust for current BMI as a possible confounding variable. Sensitivity analyses were performed with exclusion of outliers systematically identified using Cook's distance to ensure robustness of results. Exploratory subgroup analyses within each surgical group investigated bone outcomes in postmenopausal women versus premenopausal women and men. Our power calculations suggested that our sample size would allow us to detect a true difference of 0.10 g/cm² in bone density between surgical and control groups with a power of 80% and a two‐sided alpha of 0.05. Values of p ≤ .05 were considered significant, and we did not perform adjustments for multiple comparisons and therefore data should be interpreted with appropriate caveats. All analyses were performed using R version 3.4.1 (R Foundation for Statistical Computing, Vienna, Austria; https://www.r-project.org/) and R Studio version 1.1.463 (RStudio, Inc, Boston, MA, USA).

Results

Clinical characteristics

The clinical characteristics of the RYGB and AGB groups and their respective matched control groups are provided in Table 1. By design, there were no significant differences in age, BMI, sex/menopause status, or race/ethnicity between RYGB and controls, or between AGB and controls. On average, RYGB subjects received their bariatric procedure 13.7 ± 2.2 years ago, and had lost 45 ± 18 kg compared to their preoperative weight. AGB subjects received their bariatric procedure on average 11.6 ± 1.1 years ago, and reported an average weight loss of 27 ± 14 kg. One AGB patient had their band removed 10 years after their original surgery due to discomfort (17 months prior to their study visit, and with minimal weight regain after removal). Lean mass and fat mass were similar in the RYGB and AGB groups compared to their respective BMI‐matched controls, as were physical activity levels and calcium and vitamin D intake. Five RYGB subjects had a history of diabetes, although diabetes had improved and/or resolved after surgery in all. Of the four AGB subjects with a history of diabetes, one had resolution of diabetes after surgery. Although fasting glucose levels were similar to controls, the RYGB group had lower insulin levels and lower HOMA‐IR than controls, consistent with improved insulin sensitivity after bariatric surgery. There were numerically more subjects who reported a history of fractures in the RYGB group as compared to their controls (nine versus three, p = .098), which was driven primarily by an increased number of upper extremity fractures reported in the RYGB group. The AGB group had no difference in fracture history as compared to their respective controls.

Table 1

Clinical Characteristics of RYGB, AGB, and Matched Control Groups

Characteristic	RYGB	RYGB controls	p	AGB	AGB controls	p
Study subjects, n	25	25		25	25
Age (years)	56 ± 10	54 ± 12	.114	63 ± 7	62 ± 8	.685
Sex/menopause status, n			1.000			1.000
Premenopausal	9	9		2	2
Postmenopausal	13	13		20	20
Men, n	3	3		3	3
Race/ethnicity, n			.300			1.000
White	15	16		22	22
Black	9	9		2	2
Other	1	0		1	1
Surgical history
Time since surgery (years)	13.7 ± 2.2	n/a		11.6 ± 1.1	n/a
Postsurgical weight loss (kg)	45 ± 19	n/a		27 ± 14	n/a
Medical history, n
Hypertension	11	11	1.000	13	8	.252
Cardiovascular disease	1	3	.602	8	2	.077
Diabetes	5	1	.192	4	6	.724
Smoking	4	1	.346	1	0	1.000
Family history of osteoporosis	6	5	.936	7	3	.402
Prior fractures	9	3	.098	9	11	.773
Body composition
BMI (kg/m2)	33 ± 6	33 ± 6	.706	36 ± 7	35 ± 7	.481
Weight (kg)	89 ± 16	90 ± 18	.529	94 ± 21	96 ± 21	.559
Lean mass (kg)	49.6 ± 7.5	51.5 ± 8.8	.277	50.6 ± 11.8	50.8 ± 10.2	.707
Fat mass (kg)	34.5 ± 10.6	35.9 ± 10.6	.495	40.5 ± 11.3	39.9 ± 11.2	.806
Bionutrition
Physical activity (hours/week)	21 ± 19	23 ± 17	.635	21 ± 23	18 ± 14	.483
Calcium intake (mg/day)	1,138 ± 765	1,040 ± 646	.740	1,065 ± 618	1,398 ± 808	.259
Vitamin D intake (IU/day)	1,158 ± 1,358	933 ± 1,666	.461	908 ± 1,198	789 ± 719	.787
Metabolic laboratory results
Fasting glucose (mg/dL)	84 ± 8	84 ± 10	.733	90 ± 28	92 ± 15	.938
Fasting insulin (μIU/mL)	3.6 ± 2.3	6.6 ± 4.3	.004*	5.9 ± 3.7	11.7 ± 13.0	.057
HOMA‐IR	0.8 ± 0.5	1.4 ± 1.2	.007*	1.3 ± 0.8	2.7 ± 3.1	.051
Bone biomarkers
Calcium (mg/dL)	8.9 ± 0.5	8.7 ± 0.3	.087	9.0 ± 0.4	8.9 ± 0.4	.504
25OHD (ng/mL)	26 ± 9	26 ± 8	.850	33 ± 12	29 ± 9	.319
PTH (pg/mL)	61 ± 32	58 ± 26	.707	49 ± 27	45 ± 20	.488
CTX (ng/mL)	0.75 ± 0.35	0.43 ± 0.20	<.001*	0.42 ± 0.16	0.41 ± 0.22	.811
P1NP (μg/mL)	81 ± 28	51 ± 18	<.001*	58 ± 19	55 ± 13	.299
Medication usage, n
Proton pump inhibitors	5	2	.415	10	3	.053
Statins	3	6	.462	13	11	.778
Thyroid medications	4	5	1.000	3	5	.700
SSRIs or SNRIs	7	5	.741	8	6	.753

Values are mean ± SD unless otherwise noted.

HOMA‐IR = Homeostatic Model Assessment of Insulin Resistance; n/a = not applicable; SNRI = serotonin‐norepinephrine reuptake inhibitor; SSRI = selective serotonin reuptake inhibitor.

*

Value of p ≤ .05 for comparison between surgical group and their respective control group.

Biochemical markers of bone metabolism

Serum calcium, 25‐hydroxyvitamin D, and PTH were similar between the RYGB group and their BMI‐matched controls (Table 1). Of note, the number of subjects with PTH levels above the normal range was small in all groups (three in RYGB, three in RYGB‐controls; one in AGB, one in AGB‐controls). Bone turnover markers were significantly higher in subjects who had received RYGB surgery 10 years ago as compared to BMI‐matched controls, with CTX and P1NP levels that were 99% and 77% higher, respectively, in the RYGB group (p < .001 for both; Fig. 1A and B).

Fig 1

Mean ± SD of serum CTX and P1NP in RYGB, AGB, and their matched control groups. Serum CTX (A) and P1NP (B) are shown for the RYGB (black bars) versus matched control group (light gray bars), and in the AGB (dark gray bars) versus matched control group (white bars). *p value < .05 for comparison between RYGB and controls. Percent difference between surgical and control groups are shown for significant values.

Among subjects who had received AGB at least 10 years ago compared to their BMI‐matched controls, there were no significant differences in serum calcium, 25‐hydroxyvitamin D, or PTH (Table 1), nor were there differences in bone turnover markers (Fig. 1).

aBMD

Among adults who received RYGB at least 10 years ago, aBMD was on average 9.7% lower at the total hip than BMI‐matched controls and 12% lower at the femoral neck (p = .001 for both; Fig. 2A and B and Supplementary Table 1a). Bone density was not significantly different between RYGB and control groups at the spine (p = .241; Fig. 2C). We did not find any significant differences in aBMD between AGB subjects and BMI‐matched controls at any skeletal site (Fig. 2A‐C and Supplementary Table 1b).

Fig 2

Mean ± SD of aBMD by DXA in the RYGB, AGB, and their matched control groups at the total hip, femoral neck, and spine. aBMD by DXA is shown at the total hip (A), femoral neck (B), and posterior–anterior lumbar spine (C) in the RYGB (black bars) versus matched control group (light gray bars), and in the AGB (dark gray bars) versus matched control group (white bars). *p value < .05 for comparison between RYGB and controls. Percent difference between surgical and control groups are shown for significant values.

vBMD and microarchitecture

Peripheral vBMD and microarchitectural parameters showed deterioration in the RYGB group compared to their respective BMI‐matched controls at both the distal radius and distal tibia (Table 2 and Fig. 3A‐D). Total vBMD was 19% and 17% lower at the radius and tibia, respectively, in the RYGB group compared to controls (p ≤ .001 for both). These deficits appeared to be primarily driven by markedly lower trabecular vBMD at both the radius and tibia (p < .001 for both). Cortical vBMD was 7% lower at the tibia (p = .002) in the RYGB group, and was lower by a similar magnitude at the radius although this was not statistically significant (p = .067). At both the radius and tibia, trabecular microarchitectural parameters showed consistent deterioration in RYGB subjects compared to controls, with lower trabecular number, thinner trabeculae, and increased trabecular separation (p ≤ .018 for all). Cortical porosity at the tibia was greater in the RYGB group than controls (p = .006). In contrast, we found no differences in peripheral vBMD and microarchitecture at the radius or tibia between subjects who had AGB ≥10 years ago compared with their BMI‐matched controls (Supplementary Table 2).

Table 2

Bone Microarchitecture in RYGB Group and Matched Control Group

	Radius				Tibia
Parameter	RYGB	RYGB controls	Percent difference (%)	p	RYGB	RYGB controls	Percent difference (%)	p
Density
Total vBMD (mgHA/cm3)	267 ± 63	333 ± 70	−19	<.001*	254 ± 54	310 ± 47	−17	<.001*
Trabecular vBMD (mgHA/cm3)	129 ± 27	174 ± 36	−23	<.001*	140 ± 30	180 ± 28	−21	<.001*
Cortical vBMD (mgHA/cm3)	899 ± 152	958 ± 65	−7	.067	828 ± 108	901 ± 96	−7	.002*
Geometry
Total area (mm2)	301 ± 46	291 ± 49	+9	.298	763 ± 123	717 ± 116	+8	.248
Trabecular area (mm2)	245 ± 47	231 ± 50	+15	.203	632 ± 121	590 ± 119	+10	.302
Cortical area (mm2)	59 ± 13	63 ± 13	−3	.271	134 ± 25	131 ± 23	+5	.331
Microarchitecture
Trabecular thickness (mm)	0.064 ± 0.008	0.071 ± 0.012	−9	.018*	0.067 ± 0.010	0.074 ± 0.009	−10	.002*
Trabecular number (1/mm)	1.70 ± 0.35	2.05 ± 0.26	−15	.002*	1.75 ± 0.33	2.03 ± 0.27	−12	.002*
Trabecular separation (mm)	0.55 ± 0.15	0.43 ± 0.07	+31	.002*	0.52 ± 0.12	0.43 ± 0.06	+23	.001*
Inhomogeneity of network (mm)	0.31 ± 0.24	0.18 ± 0.04	+80	.018*	0.25 ± 0.09	0.19 ± 0.04	+35	.006*
Cortical thickness (mm)	0.84 ± 0.20	0.92 ± 0.21	−6	.152	1.27 ± 0.22	1.32 ± 0.28	−0	.450
Cortical porosity (%)	4.82 ± 10.97	2.19 ± 1.69	+143	.244	9.48 ± 5.42	6.38 ± 4.83	+80	.006*
Micro‐finite element analysis
Failure load (N)	3,357 ± 846	4,133 ± 924	−17	.003*	9,668 ± 1789	11,001 ± 1614	−12	<.001*

Values are mean ± SD and percent differences between RYGB and control group.HA = hydroxyapatite.

*

Value of p ≤ .05 for comparison between RYGB and control group.

Fig 3

Bone microarchitecture and trabecular morphology at the distal tibia in RYGB, AGB, and their matched control groups. (A) A representative cross‐sectional image of the distal tibia shows increased cortical porosity and trabecular microarchitectural deterioration in a RYGB subject (top panel) as compared to their age‐, sex‐ and BMI‐matched control subject (bottom panel). (B) Trabecular morphology of the distal tibia shows a lower proportion of trabecular plates in the same RYGB subject (top panel) as compared to their matched control (bottom panel). Trabecular plate volume fraction (C) and axial bone volume fraction (D) in the RYGB (black bars) versus matched control group (light gray bars), and in the AGB (dark gray bars) versus matched control group (white bars). *p value < .05 for comparison between RYGB and controls. Percent difference between surgical and control groups are shown for significant values.

Trabecular bone morphology

As assessed by ITS, trabecular bone morphology was significantly different between the RYGB group and BMI‐matched controls at both the radius and tibia (Table 3). Plates differed in number and thickness between RYGB subjects and controls to a greater degree than rods. Plate bone volume fraction was 29% and 25% lower in the RYGB group than controls at the radius and tibia, respectively (p < .001 for both). Plate number density was also lower in the RYGB group at the radius and the tibia, as was trabecular plate thickness at the tibia (p < .011 for all). Furthermore, trabecular bone in the RYGB group was less axially aligned at both the radius and tibia (p ≤ .001 for both), which could contribute to a reduced ability to sustain axial compressive forces. The AGB group, however, did not demonstrate any significant differences in trabecular bone morphology outcomes as compared to matched controls (Supplementary Table 3).

Table 3

Trabecular Bone Morphology in RYGB Group and Matched Control Group

	Radius				Tibia
Parameter	RYGB	RYGB controls	Percent difference (%)	p	RYGB	RYGB controls	Percent difference (%)	p
Bone volume fraction	0.22 ± 0.04	0.29 ± 0.05	−21	<.001*	0.25 ± 0.04	0.30 ± 0.04	−17	<.001*
Axial bone volume fraction	0.08 ± 0.02	0.11 ± 0.03	−27	<.001*	0.10 ± 0.03	0.13 ± 0.02	−19	<.001*
Plate bone volume fraction	0.07 ± 0.03	0.11 ± 0.04	−29	<.001*	0.10 ± 0.03	0.14 ± 0.03	−25	<.001*
Plate tissue fraction	0.30 ± 0.08	0.36 ± 0.10	−15	.007*	0.42 ± 0.10	0.47 ± 0.07	−11	.015*
Trabecular plate number density	1.28 ± 0.15	1.47 ± 0.17	−13	<.001*	1.42 ± 0.13	1.56 ± 0.08	−9	<.001*
Trabecular rod number density	1.83 ± 0.13	1.93 ± 0.11	−5	.012*	1.79 ± 0.16	1.86 ± 0.13	−3	.086
Mean trabecular plate thickness	0.20 ± 0.01	0.21 ± 0.01	−2	.058	0.22 ± 0.01	0.22 ± 0.01	−3	.011*
Mean trabecular rod thickness	0.21 ± 0.01	0.21 ± 0.01	−1	.435	0.21 ± 0.01	0.21 ± 0.01	−0	.598
Mean trabecular rod length	0.67 ± 0.02	0.64 ± 0.02	+4	<.001*	0.65 ± 0.02	0.64 ± 0.01	+2	.040*

Values are mean ± SD and percent differences between RYGB and control group.

*

Value of p ≤ .05 for comparison between RYGB and control group.

Estimated bone strength

μFEAs revealed a 17% lower failure load at the radius (p = .003) and 12% lower failure load at the tibia (p < .001; Fig. 4A and B) in RYGB as compared with BMI‐matched controls. In contrast, there were no significant differences in estimated bone strength between the AGB group and matched controls.

Fig 4

Mean ± SD of estimated failure load at the radius and tibia in the RYGB, AGB, and matched control groups. Estimated failure load at the distal radius (A) and tibia (B) in the RYGB (black bars) versus matched control group (light gray bars), and in the AGB (dark gray bars) versus control group (white bars). *p value < .05 for comparison between RYGB and controls. Percent difference between surgical and controls groups are shown for significant values.

Subgroup analyses

In exploratory subgroup analyses within the RYGB group, we evaluated whether there were separate patterns in postmenopausal women (n = 13) as compared with premenopausal women and men (n = 12). We found that the magnitude of bone density difference at the hip and spine was similar between the subgroup of premenopausal women and men and their respective matched controls as compared to the postmenopausal women and their matched controls (Supplementary Table 4). However, exploratory analyses did suggest that postmenopausal women had numerically greater differences from controls in estimated bone strength at the distal radius and tibia (−22% and −17% lower than controls, respectively; p ≤ .010 for both) as compared to the subgroup of premenopausal women and men (−12% and −7% for radius and tibia, respectively; nonsignificantly different from controls), although interaction tests by menopause status were not statistically significant. The differences in bone turnover markers between RYGB and controls were of greater magnitude in the premenopausal and men subgroup (CTX +139%, p = .001; P1NP +109%, p = .002) as observed in the postmenopausal subgroup (CTX +59%, p = .01; P1NP +45%, p = .04). This smaller relative increase in CTX and P1NP in the postmenopausal RYGB subgroup was likely a consequence of the overall higher levels of bone turnover seen in the postmenopausal control group as a result of estrogen deficiency.

Mediators of skeletal differences

We assessed potential mediators of skeletal outcomes separately within each surgical group. Postoperative weight loss after RYGB was negatively correlated with bone density at the total hip, femoral neck, and tibia on univariate analysis (r = −0.50 to −0.63, p < .032 for all; Supplementary Table 5). However, larger postoperative weight loss was highly correlated with lower current BMI (r = −0.52, p = .008), and adjustment for current BMI reduced all associations between weight loss and bone outcomes except for hip aBMD, which remained statistically significant (p = .020). Maximal weight loss (defined as the difference between preoperative and self‐reported nadir weight), was not associated with any bone variable. Furthermore, insulin was positively correlated with bone density at the total hip (r = 0.62, p = .003) in the RYGB group, which remained statistically significant after controlling for current BMI (p = .048). PTH was not associated with bone density outcomes in the RYGB group. Sensitivity analyses removing statistical outliers did not materially change the correlation results.

In the AGB group, current BMI, weight loss, and insulin levels were not correlated with bone density at any site in either unadjusted or adjusted analyses. There were significant negative correlations between PTH and total vBMD at the radius and tibia (p = .04 for both), but these associations were driven by the presence of a few outliers. In sensitivity analyses, exclusion of statistical outliers negated this association. Finally, P1NP and CTX levels were not significantly associated with bone density or microarchitecture outcomes in either the RYGB or AGB groups.

Discussion

This study provides the first comprehensive evaluation of bone density and microarchitecture in RYGB and AGB patients more than a decade after their procedure. We found that adults who received RYGB, but not AGB, have lower hip and peripheral bone density than controls with similar BMI, sex, age, menopause status, and race/ethnicity. Only the RYGB group had deficits in both cortical and trabecular compartments, including fewer plate‐like trabecular structures, leading to impaired estimated skeletal strength in comparison to BMI‐matched controls. Significantly higher CTX and P1NP levels in the RYGB, but not AGB, group compared to BMI‐matched controls indicate elevated bone turnover even a decade after surgery despite a lack of difference in serum 25‐hydroxyvitamin D and PTH levels. Finally, fractures were more common the RYGB group than controls.

Two prior studies have evaluated area bone density and bone turnover markers at least a decade after RYGB.^(27,28) Neither of these studies had control groups, but they did document a high proportion of negative Z‐scores after RYGB, indicating lower than expected bone density. However, by relying on DXA Z‐scores, these studies were unable to fully take into account the effect of body weight on bone outcomes. In contrast, our study compares long‐term bone outcomes to BMI‐matched controls, thus accounting for the fact that after bariatric surgery, patients tend to remain obese despite significant weight loss. Indeed, we found that RYGB subjects had notably lower bone density at the hip and peripheral sites as compared to BMI‐matched controls, but that spine bone density was similar to controls. This is consistent with longitudinal studies that document greater DXA‐assessed bone loss at the hip than the spine after RYGB.^(8,11–13) Our results suggest that bone loss after RYGB exceeds physiologic expectations for the new weight setpoint at the hip and peripheral sites, but that RYGB‐induced bone loss at the spine might be appropriate for the achieved postoperative weight. Our findings also align with cohort studies that have documented an increased risk of fracture at hip and upper extremity sites after RYGB.^(29–34)

To our knowledge, our study is also the first to investigate bone microarchitecture, estimated bone strength, and trabecular morphology in AGB patients at any point after surgery, or in RYGB patients a decade after their procedure. Similar to data suggesting that AGB has a limited skeletal impact as assessed by DXA,^(15–17) we found that more detailed investigative tools of bone microarchitecture and morphology did not show any differences in bone structure between the AGB group and BMI‐matched controls. In contrast, prior studies report significant declines in cortical and trabecular microarchitectural characteristics in the first few years after RYGB.^(8,11,13,35) Our results expand upon these data, by showing that early RYGB‐associated declines ultimately are associated with lower trabecular number, thinner trabeculae, and increased trabecular separation at the radius and tibia, as well as increased cortical porosity at the tibia as compared to BMI‐matched controls. Trabecular morphology analyses revealed a lower proportion of both rod‐like and plate‐like trabeculae in the RYGB group, with an overall greater negative impact on the generally stronger plate‐like structures and a reduction in the axial alignment of trabeculae, which result in weakened bone strength in the setting of compressive forces. Altogether, these findings contributed to a lower estimated failure load at both the radius and tibia in the RYGB subjects, and reiterate that long‐term skeletal deterioration after RYGB is associated with reduced skeletal integrity as compared to BMI‐matched controls.

Despite robust characterization of post‐RYGB bone loss, the mechanism remains unclear. One theory holds that mechanical unloading of the skeleton leads to bone loss as a physiologically‐appropriate adaption to weight loss, and some studies have shown a correlation between changes in BMD and changes in weight.^(5,12,36) In our study, we found that BMD at the hip and tibia were negatively correlated with self‐reported weight loss, although this was confounded somewhat by current BMI. Despite this correlation, bone densities at the hip, tibia, and radius sites were all lower in the RYGB group compared to BMI‐matched controls. Therefore, although mechanical unloading likely contributes to bone loss after RYGB, it cannot fully explain the extent of the observed low bone density (particularly at non‐load‐bearing sites such as the radius), nor the persistently elevated bone turnover. These findings suggest that skeletal deterioration at certain sites after RYGB exceeds the physiologic expectation of weight loss and is therefore pathophysiologic.

Additionally, it is unknown to what extent secondary hyperparathyroidism may contribute to reduced bone density in the long term. Calcium absorption decreases significantly after RYGB,^(37,38) but PTH is variably altered and may be confounded by differences in calcium and vitamin D supplementation across studies.^(8,39,40) Our study found no differences in serum calcium, 25‐hydroxyvitamin D, or PTH between the RYGB group and controls, and furthermore, the majority of our patients had PTH levels within the normal range. Over 10 years after their surgery, only a small number of bariatric subjects were adhering to the recommended postsurgical calcium intake (1200–1500 mg/day), but this did not result in suboptimal calcium levels or significant hyperparathyroidism. Last, we found no significant correlations between bone outcomes and PTH in the RYGB group.

Overall, these findings suggest other mechanisms for bone loss after RYGB. It is unknown to what extent gastrointestinal hormones and adipokines,^(3,41) inflammatory markers,⁽⁴²⁾ estrogen,⁽⁴³⁾ and/or the microbiome⁽⁴⁴⁾ contribute to skeletal deterioration in the context of metabolic weight‐loss procedures. We did observe that fasting insulin levels were positively correlated with hip bone density (independent of current BMI), consistent with the hypothesis that declining insulin levels after bariatric surgery might have a detrimental effect on the skeleton.⁽⁴⁵⁾ Further investigation into these potential mechanisms is warranted. The fact that RYGB, but not AGB, is associated with impairments in skeletal structure might be explained by the different mechanisms by which these procedures stimulate weight loss and affect physiology in both the short‐term and the long‐term. Most likely, the mechanism of bone deterioration after RYGB is multifactorial and caused by a combination of the discussed pathways.

The clinical implications of this study are important. The fact that bone density is lower than BMI‐matched controls suggests there is a pathophysiologic mechanism which leads to skeletal fragility in RYGB patients. Although we documented a relatively low incidence of T‐score ≤ −2.5 (four subjects in the RYGB group, two subjects in the AGB group), it is important to note that the average age of our RYGB cohort was only 56 years. Furthermore, obese adults fracture at higher bone density than lean controls,⁽⁴⁶⁾ and fracture risk is increased in adults who have received RYGB.^(29–34) Persistently high bone turnover rates greater than 10 years after RYGB imply ongoing bone resorptive processes, and suggest that evaluation at even later time points might reveal increased rates of osteoporosis in elderly RYGB patients. Furthermore, a prior study demonstrated that women who were postmenopausal at the time of RYGB had higher rates of bone loss than premenopausal women or men.⁽⁴⁷⁾ Although we did not have information about menopause status at the time of surgery, we observed that currently postmenopausal women in the RYGB group had a more pronounced difference from matched controls in bone strength at peripheral sites, as compared to the difference from controls observed in younger women and men. Moreover, even in the absence of osteoporotic bone density, RYGB patients have been shown to have an increased risk of fractures in comparison to both obese controls^(29,30,33) and AGB patients,^(31,32) and studies further suggest that non‐BMD‐dependent factors also play a role in fracture risk for bariatric patients.^(34,48) These findings suggest the need for studies of interventions to prevent or reduce the impact of bone loss after RYGB. Several studies have suggested that supervised exercise programs can partially mitigate bone loss after RYGB.^(49,50) It remains unknown whether other bone‐modifying treatments can fully prevent RYGB‐induced high‐turnover bone loss.

This study has limitations. Our data are cross‐sectional in nature, so we cannot conclude a causal relationship between current weight and bone density or bone quality. We are unable to determine longitudinal skeletal changes or to analyze how factors such as changing calcium and vitamin D adherence or physical activity may have played a role in bone loss in the decade between our subjects’ bariatric procedures and the present day. Nevertheless, our finding of 10% lower bone density at the hip and 99% higher markers of bone resorption in the RYGB group compared to matched controls suggest that the early observed changes in bone⁽⁸⁾ result in inappropriately low skeletal integrity for the new weight setpoint. The AGB group reported a higher use of proton pump inhibitors than controls, and therefore it is possible the bone density in the AGB group was underestimated. In the RYGB and matched control groups, however, medical comorbidities and medications were well‐balanced, and thus cannot explain the skeletal differences that we observed between these groups. Another limitation of our study is that the preoperative weight was not included in the operative note for approximately one‐half of bariatric surgery patients, and so we relied on self‐reported data for these individuals. This may have led to inaccurate assessment of weight loss, which may have impaired our power to detect significant correlations. Nevertheless, our use of BMI‐matched controls still enabled us to directly evaluate whether reductions in bone density exceeded the physiologic expectation of mechanical unloading. Our sample size was small, which may have limited our ability to detect significant differences, but unlike prior cross‐sectional studies of bariatric procedures,^(27,28) we evaluated matched controls and provided sophisticated HR‐pQCT analyses of bone microarchitecture and trabecular morphology. These data signify the importance of additional, larger studies to assess skeletal changes after bariatric surgery. Finally, direct comparisons between the RYGB and AGB groups in this study were not performed, because the two surgical groups were not matched for age and BMI.

In summary, this study provides clinically important information about bone density and microarchitecture outcomes among adults who received RYGB or AGB at least 10 years ago compared to matched controls. Our findings demonstrate that RYGB, but not AGB, is associated with lower bone density, worse bone microarchitecture and trabecular morphology, elevated bone turnover, and reduced estimated bone strength in the long‐term that exceed expectations for their lower postoperative weight. Thus, RYGB patients likely experience pathophysiologic declines in bone density in the decade after surgery, and warrant long‐term evaluation of their skeletal health.

Disclosures

Dr. Yu reports grants from NIH, and grants from Doris Duke Charitable Foundation obtained during the conduct of the study; and an investigator‐initiated research grant from Amgen Inc. that is outside the submitted work.

Acknowledgments

This work was supported by a Clinical Scientist Development Award from the Doris Duke Charitable Foundation, as well as NIH grants R03DK107869, S10 RR023405, 1UL1TR001102, and 1UL1TR002541. We thank the MGH Bone Density Center for DXA measurements, the MGH XtremeCT Core for HR‐pQCT measurements, the Brigham Research Assay Core for batch laboratory testing, and the nursing and dietary staff at the MGH Translational and Clinical Research Center for their dedicated care of the study participants.

Authors’ roles: Study design: EWY. Study conduct: EWY, KGL, MCC, and MH. Data collection and analysis: KGL, CCR, and MCC. Data interpretation: EWY, KGL, MLB, and CCR. Drafting manuscript: CCR, KGL, and EWY. Revising manuscript: EWY, MLB, and MH. Approving final version of manuscript: all authors. EWY takes responsibility for the integrity of the data analysis.

Peer Review

The peer review history for this article is available at https://publons.com/publon/10.1002/jbmr.4112.

Author notes

KGL and CCR are co‐first authors.