Risk of wrist fracture, estimated by the load-to-strength ratio, declines following sleeve gastrectomy in adolescents and young adults

Abstract

Summary

To understand whether the bone loss which occurs after vertical sleeve gastrectomy increases the risk of fracture, we used an engineering model to estimate risk in participants before and after surgery. We found that estimated risk decreased 1 year after surgery and remained lower, though had rebounded, at year 2.

Purpose

Vertical sleeve gastrectomy (VSG) improves metabolic health in young people with obesity but is accompanied by substantial loss of bone mass and estimated bone strength. We thus estimated fracture risk following VSG using the load-to-strength ratio (LSR), which integrates bone strength estimates with the predicted force of a fall.

Methods

Prospective 2-year study of youth ages 13–24 years with obesity undergoing VSG (n = 24) or lifestyle therapy (n = 34). We performed high-resolution peripheral quantitative computed tomography of the distal radius and microfinite element analysis to estimate bone strength and calculated LSR.

Results

VSG participants lost 26.4 ± 8.1% weight at year 1 (p < 0.001), which was sustained at year 2, while control participants gained weight at year 2 (4.5 ± 8.3%, p = 0.009). The predicted impact force decreased at years 1 and 2 following VSG (p < 0.001) but increased at year 2 among controls (p = 0.011). Estimated bone strength was unchanged at year 1 but decreased (p < 0.001) at year 2 following VSG, while bone strength did not change in controls. At year 1, the LSR decreased among VSG participants (p < 0.001), implying a lower risk of fracture. At year 2, the LSR was lower than baseline (p < 0.001), but higher compared to year 1 (p = 0.001). LSR did not change in the control group.

Conclusions

Short-term estimated fracture risk at the radius following VSG decreases. However, ongoing bone loss despite stable weight between years 1 and 2 leads to a concerning rise in estimated fracture risk. Longer follow-up will be critical to evaluate the trajectory of fracture risk.

(ClinicalTrials.gov NCT02557438, registered 9/23/2015).

Introduction

Youth with obesity and excess adiposity are at increased risk of fracture [1–4]. While areal bone mineral density (aBMD) as measured by dual-energy x-ray absorptiometry (DXA) is higher among youth with obesity compared to those with healthy weight [5], this may not fully compensate for the increased forces experienced during a fall due to higher body mass. Specifically, longitudinal studies suggest that bone strength in youth adapts to increases in lean mass, but not to increases in fat mass [6, 7]. In addition, volumetric bone imaging technologies have demonstrated altered bone geometry and microarchitecture in youth with excess adiposity, including decreased trabecular thickness [8–10], and excess cortical porosity [10], which may contribute to fracture risk independent of aBMD [11]. The load-to-strength ratio (LSR) is an engineering model which estimates fracture risk by comparing the predicted impact force of a fall, which depends on body mass and fall distance, to bone strength—with a higher LSR implying a greater fracture risk [12–16]. Our group has recently shown that the LSR at the distal radius is higher in adolescent and young adult females with obesity compared to healthy-weight peers, and that this difference is associated with lower percent lean mass and higher total and visceral fat mass among those with obesity [17].

Metabolic and bariatric surgery (MBS) is one of the most effective therapies for sustained weight loss in youth with obesity [18], with vertical sleeve gastrectomy (VSG) currently the most common procedure in youth [19]. In addition to weight loss, health benefits of VSG include remission of type 2 diabetes and prediabetes, and improvements in dyslipidemia, blood pressure, and renal function [20]. However, adverse skeletal consequences including bone loss and excess fracture risk are consistently reported in both adults [21] and in youth [22] following MBS. In particular, in a cohort of adolescents and young adults with moderate to severe obesity followed for 2 years, we recently reported bone mineral content loss by dual-energy x-ray absorptiometry of 4–8% depending on body site, and 6–10% loss of trabecular bone mineral density at the distal radius and tibia [23]. In addition, we have shown significant loss of vertebral bone mineral density and strength estimates in this cohort [24].

To better understand the near-term effects of VSG on fracture risk, we evaluated longitudinal changes in estimates of bone strength and the LSR at the distal radius in this cohort and compared them to a cohort in whom obesity was managed with conventional lifestyle recommendations. We hypothesized that VSG would lead to a decrease in the LSR at 2 years, given the reduction in body mass and thus predicted force of fall. In addition, given the cross-sectional correlations of lean and fat mass with LSR in youth with obesity, we investigated the relationship of VSG-induced body composition changes with changes in LSR.

Methods

Study design and participants

We conducted a two-year observational cohort study of adolescents and young adults with moderate to severe obesity to investigate the skeletal outcomes of MBS. All subjects in the present analysis either underwent VSG or received standard lifestyle guidance. The study design and two-year bone mineral density and microarchitecture results have been previously described [23]. Briefly, we recruited adolescents and young adults aged 13–24 years with an indication for MBS, either class 3 obesity (BMI ≥ 40 kg/m2 or ≥ 140% of the 9 5^th percentile for age and sex) or class 2 obesity (BMI ≥ 35 kg/m2 or ≥ 120% of the 95^th percentile for age and sex) with a relevant comorbidity [25]. Subjects were excluded if they were pregnant or breastfeeding, had untreated thyroid disease, smoked greater than 10 cigarettes/day, used glucocorticoids or anti-epileptic medications with an effect on vitamin D metabolism, or were > 450 lbs due to weight limits on the radiology equipment used in the protocol. We obtained informed consent from subjects ≥ 18 years. For pediatric subjects, we obtained informed consent from a parent or guardian and informed assent from the subject. This study was approved by the Partners Human Research Committee and was registered at ClinicalTrials.gov (NCT02557438).

Subjects were recruited between June 2015 and September 2019. 84 subjects (37 VSG and 47 control) were seen for a baseline visit, which, for the VSG subjects, occurred within 1 month prior to surgery. Follow-up study visits were conducted at 1 and 2 years after the baseline. Of the 68 subjects who completed the 2-year follow-up visit (31 VSG, 37 control), we excluded 10 who did not have a usable baseline and at least one follow-up radius HR-pQCT scan due to motion artifact, leaving 58 subjects (24 VSG and 34 control) for this analysis.

Clinical investigation

At each visit, a standardized medical history was obtained. Height was measured with a wall-mounted stadiometer and the mean was calculated from 3 separate measurements. Metabolic weight was measured to the nearest 0.1 kg with an electronic scale. Subjects self-reported their race and ethnicity. 25-hydroxyvitamin D (25OHD) was measured by immunochemiluminometric assay (LabCorp, Burlington, NC, USA).

Body composition

Body composition was measured with whole-body dual-energy x-ray absorptiometry (DXA) (Hologic 4500 A, Apex software version 13.3; Hologic Inc, Waltham, MA). The coefficients of variation for fat mass and lean mass for our institution are 2.1%, and 1.0%, respectively.

HR-pQCT

We obtained HR-pQCT imaging of the distal radius (Xtrem-eCT1, Scanco Medical AG, Brüttisellen, Switzerland) at each study visit. Scans were acquired at the non-dominant arm unless there was a history of fracture in that arm in which case the dominant arm was scanned. The region of interest was 9.02 mm long, with the distal boundary set 9.5 mm proximal to the radius endplate. Scans with significant motion artifact (≥ 4 on a 5-point scale) were excluded from analysis. For each subject, a common volume of interest (VOI) across the 3 timepoints was determined using manufacturer-provided software to match periosteal contours, and longitudinal analyses were conducted on the resultant common VOI. The mean overlap was 88%. We employed microfinite element analysis (μFEA) of the HR-pQCT images to calculated estimated failure load at the distal radius as previously described [26]. Manufacturer-provided software was used with a Young’s modulus of 10 GPa and Poisson’s ratio of 0.3. A simulated axial compression force was applied, and estimated failure load was defined as the force at which > 2% of the bone tissue experienced > 7000 μstrain [27].

Load-to-strength ratio calculation

The predicted force of impact of a fall on an outstretched arm was calculated as previously described [16]. Briefly, a single-spring model was used in which the estimated impact force depends on the height of the fall (h), the mass of the subject in kg (m), the gravitational constant (g, 9.81 m/s²) and a stiffness constant.

$$F\left(impact\right)=\sqrt{2mhgk}$$

The stiffness constant was empirically determined by Johnston and colleagues [16] to be 8989 N/m for males and 4527 N/m for females, leading to calculated forces that are highly correlated with experimentally measured impact force (R² = 0.88 for males and 0.84 for females). We modeled a fall from standing position with a fall height of half body height as described [13]. The load-to-strength ratio (LSR) was calculated as the predicted impact force divided by the estimated failure load derived from μFEA.

Statistics

Analyses were performed with Stata 12.1 (StataCorp LP, College Station, Texas). Statistical significance was defined as 2-sided p < 0.05. This was a post-hoc analysis of a larger study examining skeletal and metabolic outcomes of metabolic and bariatric surgery. Baseline characteristics between groups were compared by Student’s t-test (continuous variables) and χ² or Fisher’s exact test. Evaluations of within- and between-group changes in load-to-strength parameters were performed with linear mixed models using a group by time interaction term and adjusting for baseline age and sex. Results are reported as the estimated marginal means and standard errors. Associations of body composition and its changes with load-to-strength parameters were evaluated with linear regression, adjusted for baseline age and sex. Sensitivity analyses were performed on changes in load-to-strength parameters in the females only.

Results

Clinical characteristics

Baseline characteristics are shown in Table 1. The VSG and non-surgical subjects did not differ significantly by age, sex, self-reported race and ethnicity, or anthropometric measurements. Baseline LSR did not differ between groups. No subject in either group sustained a fracture over the course of the study. Of note, the subjects excluded for lack of an interpretable baseline or follow-up radius HR-pQCT scan were taller (172.3 ± 7.8 cm vs. 166.4 ± 8.3 cm, p = 0.040) and had higher weight (140.9 ± 36.1 kg vs. 122.5 ± 20.1 kg, p = 0.022) when compared to included subjects, but there was no significant difference in BMI between excluded and included subjects (46.9 ± 8.4 kg/m² vs. 44.2 ± 6.0 kg/m², p = 0.221).

Table 1.

Baseline characteristics

	VSG N = 24	Control N = 34	p
Age (years)	18.1 (2.2)	18.1 (2.8)	0.982
Sex (female, male)	21, 3	25, 9	0.168
Race			0.395
American Indian	0	1
Asian	0	1
Black	7	5
White	13	18
More than one	3	3
Other/not reported	1	6
Ethnicity			0.383
Hispanic	11	19
Non-Hispanic	13	15
Height (cm)	165.8 (9.3)	166.8 (7.7)	0.663
Weight (kg)	125.4 (19.5)	120.4 (20.5)	0.362
BMI (kg/m2)	45.7 (6.8)	43.1 (5.2)	0.107
Lean mass (kg)	62.4 (10.8)	61.7 (11.0)	0.829
Fat mass (kg)	61.8 (11.6)	57.5 (11.6)	0.165
Percent fat (%)	48.7 (4.3)	47.2 (4.5)	0.194
25OHD (ng/mL)	25.1 (10.2)	23.0 (8.6)	0.406
Predicted force of fall (N)	3195 (578)	3332 (814)	0.482
Estimated failure load (N)	4668 (847)	4692 (949)	0.922
LSR	0.69 (0.11)	0.73 (0.20)	0.418

Data presented as mean (SD). BMI body mass index. 25OHD 25-hydroxyvitamin D, LSR load-to-strength ratio

Two-year total body weight and body composition trajectories

Similar to the results we previously reported for the whole cohort [23, 28], subjects in the VSG group included in this analysis lost 26.4 ± 8.1% of total body weight at year 1 (p < 0.001 vs. baseline), which was sustained (−24.6 ± 10.2%, p < 0.001 vs. baseline) at year 2. Non-surgical subjects gained total body weight at year 2 (4.5 ± 8.3%, p = 0.009). Body composition analysis by DXA revealed that VSG subjects lost 15.5 ± 6.5% lean mass and 37.1 ± 14.8% fat mass at year 1 (both p < 0.001 vs. baseline) with sustained body composition changes at year 2 (−15.0 ± 6.7% lower lean mass and −33.0 ± 16.0% lower fat mass compared with baseline.) Non-surgical subjects gained lean mass (2.4 ± 4.7%, p = 0.038 at year 1 and 4.9 ± 6.1%, p < 0.001 at year 2 vs. baseline) with no significant changes in fat mass over 2 years (−4.1 ± 21.4%, p = 0.220 at year 1 and 5.9 ± 13.3%, p = 0.082 at year 2 vs. baseline).

Load-to-strength ratio trajectories

In concert with loss of weight, the predicted force of impact decreased over 2 years in all subjects in the VSG group, with an average decrease of 14.3 ± 1.1% at 1 year and 13.3 ± 1.1% at 2 years (p < 0.001 vs. baseline for both), adjusted for sex and baseline age. In the non-surgical group, we observed no change in predicted force at year 1 and a modest 2.4 ± 0.9% increase (p = 0.011 vs. baseline) at year 2 (Fig. 1A). As previously described in the larger cohort [23], estimated failure load did not change over the first year in either group, but decreased 6.9 ± 1.5% (p < 0.001) at year 2 in the VSG group (Fig. 1B). In the VSG group, the LSR decreased 12.9 ± 1.7% at year 1 (p < 0.001 vs. baseline), implying a lower likelihood of fracture. By year 2, the LSR was only 6.7 ± 1.7% lower than baseline (p < 0.001), which was significantly higher compared to 1 year (p = 0.001) (Fig. 1C). We did not observe significant changes in LSR over 2 years in the non-surgical group. In a sensitivity analysis evaluating only female participants, the results were largely similar with the exception that the increase in predicted force of fall at year 2 in control participants was no longer statistically significant (2.0 ± 0.9% increase (p = 0.091 vs. baseline).

Fig. 1.

Percent change in predicted force (A), failure load (B) and load to strength ratio (C) over 2 years. Data are marginal means, adjusted for sex and baseline age, for VSG group (filled squares) and controls (open squares). * p < 0.05 compared to baseline. † p < 0.05 between groups. a p < 0.05 compared to Year 1. VSG, vertical sleeve gastrectomy

Body composition and load-to-strength ratio

We next investigated the associations of body composition trajectories with changes in the LSR. As we have shown in the full cohort [17], at baseline, adjusted for age and sex, lean mass positively correlated with both the predicted force of fall (β = 0.47, p < 0.001) and with estimated failure load (β = 0.65, p < 0.001), while there was no correlation with LSR (β = −0.12, p = 0.411). Fat mass correlated with predicted force of fall (β = 0.36, p < 0.001), but did not correlate with estimated failure load (β = 0.07, p = 0.606). There was a positive correlation of fat mass with LSR (β = 0.27, p = 0.009), implying an increase in fracture likelihood with increasing adiposity (Table 2). In a sensitivity analysis evaluating females only, results were similar, except that the association of fat mass with LSR was only marginally statistically significant with a similar beta estimate (β = 0.28, p = 0.050).

Table 2.

Associations of body composition with predicted force of fall, failure load, and LSR at baseline in the entire cohort

	Predicted force of fall		Failure load		LSR
	β	p	β	p	β	p
Total weight	0.41	< 0.001	0.28	0.043	0.15	0.206
Fat mass	0.36	< 0.001	0.07	0.606	0.27	0.009
Lean mass	0.47	< 0.001	0.65	< 0.001	−0.12	0.411

Models adjusted for age and sex. β, standardized coefficient

Statistically significant values are indicated in bold font

Among subjects who underwent VSG and who had interpretable radius scans at both baseline and 2 years (n = 18), the percent weight loss, percent lean mass loss, and percent fat mass loss correlated with decreases in the predicted force of fall (Table 3). Conversely, as we have previously shown [23], there was no association of loss of weight, lean mass, or fat mass with the observed decreases in failure load. Decreases in LSR were thus associated with decreases in weight and lean mass, reflecting a lower fall force, and were borderline significantly associated with decreases in fat mass as well. When limiting the analyses to females, we did observe a significant association of decrease in failure load with decrease in fat mass (β = 0.55, p = 0.035) as well as a marginal association with decrease in total body weight (β = 0.51, p = 0.053).

Table 3.

Associations of 2-year changes in body composition with 2-year changes in LSR parameters in the VSG group

	%Δ Predicted force of fall		%Δ Failure load		%Δ LSR
	β	p	β	p	β	p
%Δ Total weight	1.00	< 0.001	0.27	0.310	0.49	0.041
%Δ Fat mass	0.97	< 0.001	0.27	0.313	0.47	0.052
%Δ Lean mass	0.96	< 0.001	0.11	0.719	0.60	0.022

Models adjusted for age and sex. β, standardized coefficient

Statistically significant values are indicated in bold font

Discussion

The risk of fracture depends on a complex interplay of fall risk, the load sustained during a fall, and mechanical integrity of the skeleton. Youth with obesity are at increased risk of fracture, with many potential contributing factors including altered fall mechanics, greater force of impact during a fall, and negative effects of adipose tissue on bone mass and accrual [29]. While VSG is an effective tool to treat obesity, the associated bone loss may further exacerbate this elevated fracture risk. However, our data using LSR at the distal radius show a decline in estimated fracture risk at 1 year following VSG, driven by a decrease in weight leading to a decrease in the predicted force of fall. As bone loss progresses and estimated failure load decreases at year 2, the LSR increases, though it remains lower than at baseline. These data are consistent with some large database studies in adults, which suggest that VSG does not increase fracture risk, and may even reduce risk of fracture at certain anatomic sites [30, 31]. These data contrast with findings in adults having undergone Roux-en-Y gastric bypass (RYGB), which consistently demonstrate 1.3–2.3 fold increases in fracture risk [32–35]. It is important to interpret these data with caution however, given the overall small numbers of VSG recipients in the published fracture studies to date [21].

Our data appear to be reassuring that, despite substantial bone loss and decreases in estimated strength (as demonstrated by lower failure load), fracture risk at the radius is, if anything lower than the baseline risk in the first 2 years following VSG in young people. However, the trajectory of LSR and the significant rise in LSR between years 1 and 2 following surgery raise concern that ongoing bone loss without further reductions in body weight may eventually increase fracture risk. Long-term studies of adults undergoing RYGB demonstrate ongoing bone loss after the first 2 post-surgical years, associated with persistent elevations of bone turnover markers [36, 37]. In addition, a cohort of patients 10 years following RYGB had lower BMD compared to weight-matched controls, with c-terminal telopeptide of type 1 collagen (CTX), a marker of bone resorption, twice as high as that of controls [38]. While the increase in bone resorption and degree of post-surgical bone loss following VSG may be lower than that following RYGB [30, 39], these data suggest that longer-term follow-up, particularly in this younger population, will be critical.

The adverse effects of VSG on bone strength are particularly concerning in adolescents and young adults, most of whom are not yet at peak bone mass. DXA-derived areal BMD measures from the Bone Mineral Density in Childhood Study indicate ongoing acquisition of bone mineral content in late adolescence in both boys and girls, particularly at the distal 1/3 radius, where accrual continues into the 3^rd decade of life [40]. HRpQCT volumetric imaging demonstrates that much of this late adolescent bone accrual is of cortical bone – due both to increasing cortical thickness and decreasing cortical porosity [41]. Peak bone mass is likely a major determinant of lifelong fracture risk; data both from studies of mother-daughter dyads as well as computer modeling suggest that peak bone mass exerts a greater influence on osteoporotic risk than does rate of bone loss with aging and menopause [42–44]. These considerations temper interpretation of the favorable effect of VSG on estimated fracture risk and underscore the need for longer-term follow-up, both of skeletal phenotype and of fracture outcomes.

Evaluation of effects of body composition changes indicated that, though higher lean mass was associated with higher estimates of bone strength at baseline, loss of lean mass substantially lowered the predicted force of fall and was thus correlated with the observed decrease in LSR, implying lower fracture risk, following VSG. The role of muscle in promoting bone health is likely mediated by many pathways, including promoting anabolic activity via mechanical loading [45] as well as via paracrine action of muscle-secreted factors such as IGF-1, FGF2, irisin, and β-aminoisobutyric acid [46–48]. Conversely, certain muscle-derived factors may promote bone resorption, including interleukin-6 and myostatin [48]. These and other pathways may underlie our finding of a baseline association of lean mass with estimated failure load. Conversely, rapid changes in muscle metabolism and total mass may have effects on bone mass distinct from those observed under relatively steady conditions, leading to a lack of association of changes in lean mass with changes in estimated failure load.

Limitations of this study include the relatively small size of our cohort, though, given under-utilization of MBS in pediatric patients meeting surgical criteria [19], our study represents one of only few longitudinal cohorts of post-surgical patients in this age group [49, 50]. We used the LSR to model fracture risk rather than directly assessing post-surgical fracture incidence as substantially larger cohorts would be required given the rarity of this outcome in this age group. This ratio, also called “factor-of-risk”, has been validated in a case–control study of post-menopausal women, in which the LSR at the distal radius was associated with fracture risk [51]. Similarly, the LSR at the hip in a sideways fall configuration was a significant predictor of fracture in older adults in the Framingham Study [15]. Of note, validation studies in children have not been performed to date. In addition, the calculation of predicted force of fall relies on a stiffness constant empirically derived in healthy young adults [16]; the stiffness constant in younger people as well as in those with obesity may differ. This would be expected to alter the absolute value of the LSR but would be less likely to affect the trajectories of LSR reported here. Finally, our analyses apply to fracture risk only of the distal radius in a fall on an outstretched hand and we cannot generalize to fractures at other sites.

In conclusion, we have shown that the estimated strength at the distal radius decreases over 2 years following VSG; however, significant decreases in total body weight and thus the predicted force of a fall lead to a lowering of the LSR and thus a reduction in the estimated fracture risk. Analysis of body composition changes suggests that despite the generally positive associations of lean mass with bone parameters, loss of lean mass following VSG is not associated with worsening of estimates of bone strength and fracture risk. However, the J-shaped trajectory of the LSR in the first 2 years following VSG and the concern that bone loss may progress in subsequent years highlights the need for longer-term follow-up to better understand the implications of MBS on bone health in young people.

Data Availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.