3.2. Biological Parameters

All the biological parameter variations are presented in Table 1 and Figure 2. No initial calcium or phosphorus variation occurred, whereas calcium was modestly decreased at 12 and 24 months (−0.9 and −1.4%, respectively, p < 0.001 for both) and phosphorus transiently increased at 12 months (11.3%, p < 0.001). Serum iPTH and albumin levels remained relatively stable throughout the 24-month period, whereas 25(OH)D3 levels were significantly increased at 12 and 24 months (p < 0.001) after surgery. The prevalence of patients presenting 25(OH)D3 deficiency was 39.8% before surgery and 37.5% at 1 month, 14.9% at 12 months and 18.3% at 24 months after surgery.

Open in a separate window

Figure 2

Variation in biological parameters: calcium (A), phosphorus (B), iPTH (C), 25(OH)D3 (D), Albumin (E), osteocalcin (F), CTx (G), Periostin (H), sclerostin (I) and semaphoring (J) at different time points (baseline, 1, 12 and 24 months) following sleeve gastrectomy. iPTH: intact parathyroid hormone; 25(OH)2D3: 25 vitamin D; CTX: type I-C telopeptide breakdown products. Lower whisker represents the smallest observation ≥ lower hinge − 1.5 × IQR. Upper whisker represents the largest observation ≤ upper hinge + 1.5 × IQR. Lower and upper hinges represent the 25 and the 75% quartile values, respectively. Open circles represent the mean value at each time point. Data beyond the end of the whiskers are called outlier points and are plotted individually. IQR: interquartile range; * indicates a significant variation for p < 0.05, ** for p < 0.01, and *** for p < 0.001.

Osteocalcin and especially CTX values increased after 1 month and peaked at 12 months. From 12 to 24 months, these two bone turnover markers decreased, but the values did not return to baseline levels. Although these two parameters presented similar kinetic profiles, the magnitude of difference compared to baseline values was higher for CTX than for osteocalcin. At 1 month, periostin and sclerotin values showed a modest but significant 12% increase (p < 0.001 for both) but had returned to basal values after 12 months. Sclerostin levels continued to decrease after 12 months to reach a value lower than baseline at 24 months (p < 0.010). No initial variation in SEMA4D levels was observed at 1 month, although a modest but significant decrease (10.5% and 11.8%, respectively) occurred after 12 and 24 months (p < 0.001 for both).

3.3. Areal Bone Mineral Density

The baseline and changes in aBMD values at 12 and 24 months after SG are presented in Table 2 and Figure 3. At baseline and for all bone sites, patients with obesity presented Z-score values higher than zero, indicating that the aBMD was above the age- and gender-related reference values. This bone adaptation was particularly marked at total hip (1.09 ± 0.97 SD) and at one-third radius (2.11 ± 1.45 SD).

Open in a separate window

Figure 3

Variation in areal bone mineral density (g/cm²) in whole body (A), total hip (B), femoral neck (C), lumbar spine (D) and radius (E) at different times (baseline, 12 and 24 months) measured at whole body and various localised bone sites. Lower whisker represents the smallest observation ≥ lower hinge − 1.5 × IQR. Upper whisker represents the largest observation ≤ upper hinge + 1.5 × IQR. Lower and upper hinges represent, respectively, the 25 and 75% quartile values. Open circles represent the mean value at each time point. Data beyond the end of the whiskers are called outlier points and are plotted individually. IQR: interquartile range; ** indicate a significant variation for p < 0.01, and *** for p < 0.001.

Table 2

Areal bone mineral density of the patients at baseline, 12 and 24 months after sleeve gastrectomy.

	Baseline (n = 83)	12 Months (n = 76)	24 Months (n = 60)	Relative % Variation (Δ 12m-Baseline/Baseline)	Relative % Variation (Δ 24m-Baseline/Baseline)
Whole body
aBMD (g/cm2)	1.187 ± 0.098	1.225 ± 0.110	1.176 ± 0.113	2.4 ± 2.9 ***	−2.4 ± 5.4 **
T-score (SD)	0.71 ± 1.02	1.14 ± 1.02	0.54 ± 1.24
Z-score (SD)	0.63 ± 0.87	0.99 ± 0.92	0.55 ± 1.05
Total hip
aBMD (g/cm2)	1.083 ± 0.133	1.020 ± 0.137	1.017 ± 0.133	−6.6 ± 4.2 ***	−7.9 ± 4.6 ***
T-score (SD)	0.84 ± 0.97	0.43 ± 0.93	0.36 ± 0.90
Z-score (SD)	1.09 ± 0.97	0.66 ± 0.96	0.69 ± 0.91
Osteoporosis n (%)	0 (0%)	0 (0%)	0 (0%)
Osteopenia n (%)	2 (2.4%)	5 (6.6%)	5 (8.3%)
Femoral neck
aBMD (g/cm2)	0.920 ± 0.150	0.874 ± 0.141	0.871 ± 0.140	−5.4 ± 7.0 ***	−5.8 ± 7.6 ***
Lumbar spine
aBMD (g/cm2)	1.087 ± 0.139	1.095 ± 0.142	1.107 ± 0.159	−0.1 ± 4.1	−0.4 ± 5.0
T-score (SD)	0.58 ± 1.31	0.63 ± 1.30	0.71 ± 1.44
Z-score (SD)	0.86 ± 1.27	0.92 ± 1.31	1.11 ± 1.4
Osteoporosis n (%)	1 (1.2%)	0 (0%)	1 (1.7%)
Osteopenia n (%)	9 (10.8%)	8 (10.5%)	8 (13.3%)
Radius
aBMD (g/cm2)	0.641 ± 0.061	0.638 ± 0.061	0.632 ± 0.069	−0.8 ± 1.6 **	−2.1 ± 2.4 ***
T-score (SD)	1.71 ± 1.36	1.67 ± 1.25	1.43 ± 1.41
Z-score (SD)	2.11 ± 1.45	2.08 ± 1.40	1.95 ± 1.54

Open in a separate window

Legend. Data are presented as mean ± SD (standard deviation). aBMD: areal bone mineral density. The T-score describes the number of SDs by which aBMD differs from the mean value expected in young healthy individuals of the same sex, whereas the Z score describes the number of SDs by which aBMD differs from the mean value expected for age and sex. Measure Δ 12m-baseline/baseline represents relative % difference between values at 12 months and baseline. Measure Δ 24m-baseline/baseline represents relative % difference between values at 24 months and baseline. Relative % variation was defined as [100 × (measure 2 − measure 1)/measure 1]; ** indicate a significant variation for p < 0.01, and *** for p < 0.001.

After 12 and 24 months, a relative stability of aBMD was observed at whole body, lumbar spine and one-third radius, whereas a significant decrease was observed at femoral neck (−5.4% and −5.8%) and total hip (−6.6% and −7.9%). Nevertheless, at total hip, the % relative variation ranged from −19.5 to 2.9% at 12 months and −19.4 to 2.9% at 24 months, both denoting large inter-individual variations. After 12 and 24 months, mean Z-scores remained above zero for all bone sites. Using the WHO classification, 2.4% and 10.8% of patients were diagnosed with osteopenia at baseline when total hip or lumbar spine were used for the diagnosis, whereas, respectively, none and one was diagnosed with osteoporosis. At lumbar spine, the prevalence of osteopenia or osteoporosis did not vary postoperatively, but at total hip, the prevalence of osteopenia increased modestly by 6.6% and 8.3% at 12 and 24 months, respectively.

4.1. aBMD

As suggested by Matos et al. [36], BS provides a unique condition in which an extreme variance from high to reduced body-weight-bearing is perceived by the individual in a very short period. Consequently, biomechanical effects should occur mainly in bone areas submitted to higher load bearing. In accordance with this assumption, at 12 months, a noticeable and significant decrease in aBMD was mainly observed at femoral neck (−5.4%) and total hip aBMD (−6.6%), while at whole body, lumbar spine and radius, the aBMD variations were within the precision error of the DXA measurement. Between 12 and 24 months, the loss continued at femoral neck and total hip, but the decrease was slowed down compared to the first 12 months (−5.8% and −7.9%, respectively). Using a similar surgical procedure, Hofso et al. [16] reported a similar aBMD loss around 5% at femoral neck and 7.8% at total hip, while at lumbar spine and whole body, minor variations occurred after 12 months. A recently published meta-analysis including 22 studies [37] noted an obvious aBMD loss at femoral neck and none at lumbar spine. This conclusion should be considered with caution, as some of the included studies reported an aBMD decrease at lumbar spine [21]. The type of population under study or the degree of weight loss may partially explain these divergent results. Moreover, a recent meta-analysis underlined that, having QCT and HR-pQCT as a reference, DXA was found to significantly underestimate lumbar spine post-RYGB aBMD losses [38].

To our knowledge, although the analysis of non-weight-bearing bone (i.e., radius) may improve our understanding of the potential effects of endocrine modifications independently of mechanical loading, no data on aBMD variations following SG are available. After 12 and 24 months, we observed a limited decrease in aBMD, around 2% at radius, but the Z-score remained much higher compared to the reference values (+1.95 SD). Using another technique (i.e., HR-pQCT) and surgical procedure (i.e., RYGB), Shanbhogue et al. [15] also reported a minimal variation in volumetric BMD (vBMD) at the trabecular and cortical compartments of the radius after 12 months, while at 24 months, a decline in total vBMD occurred, mainly due to an alteration in the trabecular compartment [15]. This suggests that an energy deficit and changes in adipose or gastrointestinal hormones could be additional contributors to BS-induced bone loss [39], but probably to a smaller extent than the reduction in mechanical loading.

In fact, although a noticeable aBMD decrease occurred after SG at the hip region, our data do not conclusively support an increased prevalence of osteopenia or osteoporosis over the 24 months. Several studies also reported a very low incidence of T-scores ≤ −2.5 SD according to the standard DXA definition [40,41]. In our study, it is further interesting to note that the bone loss, despite being low, continued after 12 months, whereas body weight was already stabilised. This finding supports the hypothesis that aBMD loss and fracture risk are not dependent only on weight loss and rather increase with observation time [15,42].

4.2. Bone Markers

The use of biochemical markers of bone turnover may have helped to elucidate the mechanisms involved in this bone loss. After SG, the variations in bone formation and bone resorption seem to have the same kinetic profile characterised by an initial increase from the third month, a peak at around 12 months and a decrease thereafter, but with values at 24 months always remaining higher than the basal values. Nevertheless, the bone formation marker (i.e., osteocalcin) level increased to a lesser extent than the bone resorption marker (i.e., CTX), suggesting an uncoupling of bone cell activities in favour of bone resorption. This process may explain the reduction in aBMD in the first 24 months. Moreover, the persistence of a high degree of bone turnover at the end of the follow-up may also suggest that aBMD loss is likely to continue for a longer time after BS and that aBMD monitoring by DXA should thus be performed beyond 24 months in these patients. Our results fully confirmed the recently published results of Paccou et al. [17], who studied a small group of 32 patients of both sexes and women with different menopausal status and found a similar profile of variation in bone turnover using procollagen type I N-terminal propeptide (PINP), another bone formation marker, and CTX after 12 and 24 months. Muschitz et al. [21] performed a longitudinal study of 52 premenopausal women, using multiple evaluation points, and also showed an ongoing increase in CTX and PINP from 1 month to 12 months, and these values remained elevated at 24 months. Lastly, Hofso et al. [16] studied obese patients of both sexes with T2D and women with different menopausal status and showed a progressive increase in PINP and CTX from 5 to 52 weeks after SG. Up to now, it has not been clear whether the type of surgery has specific effects on bone marker levels, with some studies reporting a greater detrimental effect of RYGB compared to SG [16,17] and others finding no difference [21].

Numerous factors may influence bone cell activities, including mechanical and endocrine factors. It was suggested that the increase in bone turnover following BS may be related to the increase in calcium and parathyroid hormone [43]. However, as also previously observed, we reported no variation in either of these biological parameters [15,21,39], even though pre-surgical and post-surgical elevated PTH was also reported [21,43]. At baseline, the obese patients presented lower mean values of vitamin D (<30 ng/mL) and, although the values increased with duration after surgery, they remained at the lower limits [33]. As observed in our study, hypovitaminosis D is frequently reported in obese patients, possibly partially due to sequestration of vitamin D in subcutaneous and visceral fat [44]. It has been hypothesized that the reduction in adipose tissue, which releases stored vitamin D into the circulation associated with high vitamin D supplementation [15,39], may partially explain the increase in vitamin D but strongly suggests the need for adapted vitamin D supplementation. It was reported that 39% of patients presented postoperative vitamin D deficiency despite daily multivitamin supplementation after a restrictive procedure such as SG [45].

Conversely to calciotropic hormones and the routinely used bone turnover markers, periostin, sclerostin and SEMA4D have been poorly or never evaluated after SG [17,21,39]. We reported for the first time an initial, transient and proportional increase in periostin and sclerostin at 1 month. However, over the first 12 months, and although OC and CTX remained elevated, periostin and sclerostin returned to basal values, suggesting their limited effect on bone cell differentiation and activities. A similar conclusion could be drawn after 24 months, where a modest decrease was observed for sclerostin alone. Our results concerning sclerostin fully confirm those of Paccou et al. [17] at 12 and 24 months, even though unfortunately no investigations in the acute phase were available in their study. In premenopausal women, Muschitz et al. [21] reported that sclerostin increased from 1 month after BS, peaked at 6 months and remained higher than baseline up to 24 months, although the value decreased continuously. An increase in sclerostin was also observed in adolescents 12 months after SG [39]. The inconsistent results for sclerostin at 12 and 24 months following BS may be related to patient characteristics or the intensity of body-weight loss, but they seem to be independent of the type of surgery [17,21]. The acute reduction in mechanical loading induced by a 10 kg weight loss [21] associated with the limited daily physical activity [46] during the first month following BS may explain the increase in sclerostin levels. This assumption is based on experimental studies that demonstrated that sclerostin expression is upregulated under conditions of mechanical unloading and downregulated by mechanical stimulation [47]. However, the involvement of sclerostin in post-surgical bone loss after BS remains questionable because its physiological action is to suppress osteoblast activity [48], which has not been currently or previously observed [16,17,21].

In parallel, an unexpected transient increase in periostin levels at 1 month was also observed. In fact, experimental studies demonstrated that a reduction in mechanical loading is associated with a concomitant deterioration in bone structure and a decrease in periostin gene expression [49]. Conversely, an increase in mechanical constraints using an axial compression load on mouse tibia induced the overexpression of periostin and the downregulation of sclerostin [23]. In humans, the effect of mechanical loading or unloading on circulating periostin levels remains more controversial [50,51]. However, in the same unexpected way as after SG, we also reported a higher level of periostin in persons with spinal cord injury during the acute phase of immobilisation, when the bone loss is maximal, compared to the value in the chronic phase, when the bone loss is reduced [50]. We [50] and others [52] interpreted this high periostin level as a “protective mechanism” or as a “compensatory mechanism” in the case of low bone mass or during an acute bone loss phase. The transient increase in periostin levels at 1 month—a period when weight loss and thus the variation in mechanical loading are maximal—could go in this direction. Nevertheless, the normalisation of sclerostin and periostin levels with the post-surgery duration, whereas bone turnover remains intense, argues for a limiting effect of periostin/sclerostin on induced SG-related bone loss.

From a more exploratory perspective, we report the SEMA4D variations after SG for the first time and note that the kinetics were characterised by a weak but significant decrease only at 12 and 24 months. SEMA4D is produced by osteoclasts and suppresses the differentiation and activity of osteoblasts [30,53], which does not coincide with the higher post-surgery bone formation values compared with baseline. We must, however, bear in mind that the role of this transmembrane protein in human bone has been little investigated. In one study, a higher serum SEMA4D level was observed in postmenopausal osteoporosis patients compared to healthy controls, and the concentrations were negatively correlated with lumbar spine aBMD [29], bone alkaline phosphatase and bone Gla-protein and positively correlated with markers of bone resorption (TRACP-5b and NTX) [29]. Similarly, Anastasilakis et al. [54] reported a trend in postmenopausal women towards a negative correlation between SEMA4D levels and aBMD measured at lumbar spine but not at femoral neck.

4.3. Factors Influencing aBMD Loss

Although the bone loss appeared limited in terms of localisation (i.e., hip region) and intensity (~8%), a noticeable inter-individual variability was observed, which might be of help in identifying the factors influencing aBMD after SG. This may also be of clinical interest because fracture risk seems to increase in patients after BS [18]. Multivariate analysis clearly showed that weight change appeared to be the main parameter influencing aBMD. Interestingly, this effect was observed at weight-bearing (i.e., hip) and non-weight-bearing (lumbar spine and radius) bone sites, suggesting a systemic effect of weight loss that does not appear to be mediated by sclerostin, periostin or SEMA4D. Age, gender and most of the biological parameters (iPTH and osteocalcin, except for a few sites and time points) were also not associated with the aBMD variations. Previous studies have also found that weight change correlated with femoral neck, total hip, lumbar spine and whole body aBMD loss [21,55,56,57]. However, other studies have observed that the effect of weight loss on aBMD loss was not independent of the effect of type of surgery [16]. These findings indicate that gastric bypass per se elicits bone turnover and induces bone loss, and they do not support the idea that reduced aBMD after gastric bypass is solely a physiological adaption to a lower body weight [16].

Taking into account the potential deleterious effects of weight loss, which reduces the mechanical loading applied on the skeleton, it will be interesting to evaluate strategies for restoring bone loading. In addition, previous clinical trials have demonstrated that physical exercise, combined or not with adequate supplementation (vitamin D, calcium and protein), is a practical approach to reduce the bone loss caused by bariatric surgery [58,59]. Other studies are necessary to demonstrate whether this approach should be generalized for all patients undergoing bariatric surgery. The optimal components of physical exercise (frequency, intensity, type and duration) to improve these effects on bone health will also need to be determined, knowing that a low-dose exercise program had no favourable impact [60].

We express special gratitude to the CHU Montpellier, France, for funding this study. We are thankful to the technicians and nurses from the Departments of Clinical Physiology, Endocrinology and Nuclear Medicine for their support of this research study. Lastly, we are thankful to the study participants without whom this research endeavour would not have been possible.