This is a post hoc analysis of a prospective bariatric surgery cohort conducted at the bariatric center of Shanghai Jiao Tong University school of medicine Affiliated Sixth People’s Hospital, China. We recruited patients with obesity (women aged < 45 years with regular menstrual cycles and men aged < 50 years) who underwent SG per Chinese guidelines criteria and completed the comprehensive medical examination preoperatively (18). The exclusion criteria were as follows: (1) bone diseases of different etiologies (primary hyperparathyroidism, Paget disease, clinical or laboratory evidence of hepatic or renal failure, previous BS, or excessive alcohol intake (> 3 U/d)); (2) use of medications affecting bone metabolism, including bisphosphonates or teriparatide (in the last year or for > 12 months), oral glucocorticoids, insulin, thiazolidinediones, calcium, and vitamin D supplementations; (3) specific endocrine diseases that could cause obesity, such as hypothyroidism, polycystic ovarian syndrome and pituitary tumor; (4) missing data on age, sex, body mass index (BMI), or laboratory investigations; (5) gestation or lactation or malignant disease; (6) missing the 12 month’s follow-up visit.

All patients were prospectively examined at baseline and 6, 12 months postoperatively. 50 μg of vitamin D3 and 600 mg of calcium were prescribed as an oral supplement in all patients postoperatively daily (19). All participants signed informed consent. This study has been registered in the Chinese Clinical Trial Registry (ChiCTR2000037647).

Laparoscopic SG involved removing a major portion of the gastric greater curvature. A 32-36 French oral gastric tube was placed through the pylorus into the duodenal and left near the lesser curve. The dissection was started 2–6 cm from the pylorus and straight toward the cardiac notch angle of His over the 32-36 French bougie. The total fundus was removed with the antrum partially retained. Two metabolic surgeons performed the SG (20) operations for all patients using standardized laparoscopic techniques.

Demographic and health condition data, including age, sex, medical history, were recorded preoperatively. Anthropometric variables were measured at each visit. Details of recording the anthropometric measurements, including height (cm), weight (kg), BMI (kg/m²), waist circumference, systolic blood pressure, and diastolic blood pressure, have been described previously (19). The excess weight loss percent (%EWL) was calculated as follows: 100× (preoperative weight−current weight)/(preoperative weight−ideal weight); BMI of 24 kg/m² was used to calculate the ideal weight (21). The percent of total weight loss (%TWL) was calculated as 100 × (initial weight - postoperative weight)/initial weight. The estimated body fat percent (%eBF) was calculated as follows: (1.20 × BMI) + (0.23 × age) − (10.8 × sex) − 5.4, with 1 used for men and 0 for women. The estimated fat-free mass index (eFFMI) was calculated as follows: BMI × (1− %eBF/100) (22).

Biochemical variables were measured at each visit. A 12-h overnight fasting blood sample was collected for all participants to determine the bone metabolism, lipid profile, fasting plasma glucose (FPG), fasting insulin, glycated hemoglobin (HbA1c), total serum protein (TSP) and serum albumin (SA). Homeostatic Model Assessment for Insulin Resistance (HOMA-IR) was assessed as fasting insulin (mIU/L) × fasting glucose (mM)/22.5. HOMA-β was calculated as follows: (20 × insulin [mIU/mL])/(glucose [mmol/L]– 3.5), as previously described (23). Bone metabolism parameters, including parathyroid hormone (PTH), 25-hydroxy vitamin D (25[OH]D), procollagen type I amino-terminal propeptide (P1NP), N-terminal osteocalcin (N-MID), and C-terminal telopeptide of type 1 collagen (β-CTX) were measured using an automated Roche electrochemiluminescence system (Roche Diagnostics GmbH, Germany).

Areal BMD (g/cm²) was measured using DEXA preoperatively and at 6 and 12 months postoperatively. BMD data of the lumbar spine (LS1–4), femoral neck (FN), and total hip (TH) were recorded. All DEXA scans were performed using Hologic QDR-2000 (Hologic Corporation, Waltham, MA, USA) by two experienced operators following standard manufacturer protocols, as described previously in detail (19). Daily measurement of the Hologic DEXA spine phantom was performed for quality control maintenance. The coefficient of variation for BMD was 0.8–1.0%. Prodigy encore software (version 6.70, standard-array mode; GE Healthcare) was used for data analysis.

Body composition parameters, including VFA, SFA, and PMA, were measured using Achieva 3.0T MRI system (Philips Healthcare, Eindhoven, Netherlands). A medically trained technician analyzed a 10-mm slice at the L3 level with good contrast using sliceOmatic software (version 5.0; TomoVision, Magog, Canada). The software calculates the areas of different tissues expressed as cm², as previously reported (24).

The Shapiro–Wilk test was used to evaluate the normality of all continuous variables. Variables with a normal distribution were presented as means ± standard deviations, while the skewed variables were expressed as medians (interquartile ranges). A paired-sample t-test or Wilcoxon signed-rank test was used to compare the changes in the clinical characteristics at baseline and 6 or 12 months after BS. GLMM (generalized linear mixed model) was used to evaluate time × sex interactions for participants undergoing SG. With adjustment for baseline values, the percent change from baseline to 6 months or 12 months for the parameters was calculated as follows: 100 × (initial measurement–postoperative measurement)/initial measurement. Pearson or Spearman analysis was used to assess the correlations between BMD and clinical variables including body composition parameters. Multiple stepwise regression analysis was used to estimate the influence of different body composition parameters on BMD at different sites. The results were presented as adjusted regression coefficient (β) values and adjusted coefficient of determination (R²). All analyses were 2-tailed with a significance level (alpha) of 0.05. All data analyses were performed using IBM Statistical Product and Service Solutions Statistics for Windows, version 26.0 (IBM Corp, Armonk, NY, USA). All graphs were made using Graphpad Prism 8.0 (GraphPad Software, San Diego, CA, USA). The post hoc power analysis was calculated using Power Analysis and Sample Size 2021 (NCSS, USA).

A total of 41 participants (25 non-menopausal adult women and 16 men) were invited. The mean age was 31.62 ± 6.72 years with no significant difference between genders. The changes in metabolic indicators are shown in Table 1 . As expected, most of the glucose–lipid metabolic parameters including total glycerides (TG), low-density lipoprotein cholesterol (LDL-c), FPG, HbA1c reduced significantly from baseline to 12 months postoperatively (P < 0.05). As for factors related to bone metabolism, 25[OH]D increased significantly in both genders from baseline to 6 months postoperatively, while calcium level showed no significant changes in both genders, although these patients took vitamin D3 and calcium postoperatively daily. Additionally, PTH decreased significantly at 12 months after surgery in women while significantly decreased 6 months after surgery in men.

The changes in body composition parameters are shown in Table 2 . The mean BMI at baseline was 35.05 ± 5.06 kg/m² with no significant difference between genders. At 12 months postoperatively, the mean reduction in the %TWL and %EWL was 29.13% and 90% in women, and 26.95% and 91% in men, respectively. In the first 6 months, significant decreases in VFA, SFA, PMA and eFFMI values were observed regardless of gender, while only female showed a continued decline for 12 months. Across sex, men demonstrated greater VFA reduction than women at 6 months. eFFMI in men declined more than in women at both 6 and 12 months. Sex-specific responses of SG for body composition parameters were shown in GLMMs ( Table 3 ). The post hoc statistical power (%) of VFA, SFA, and eFFMI in paired-sample t-test was 99.82%, 100%, and 100% for women group; 98.62%, 99.95% and 100% for men group, respectively.

In our cohort, no cases of osteoporosis or fractures were reported, while clinically significant reductions in the BMD (> 0.03 g/cm²) at one or more sites (25) were observed in 36 patients (87.80%). The changes in bone mineral density after sleeve gastrectomy are shown in Figure 1 . For women, the mean FN BMD (1.045 ± 0.133 vs. 1.004 ± 0.127 g/cm²; P < 0.001) and TH BMD (1.121 ± 0.149 vs. 1.060 ± 0.143 g/cm²; P < 0.001) decreased significantly across the 12 months. The same downward trend was observed in men. The mean FN BMD (1.036 ± 0.131 vs. 0.972 ± 0.117 g/cm²; P < 0.001) and TH BMD (1.049 ± 0.125 vs. 0.957 ± 0.107 g/cm²; P < 0.001) decreased significantly across the 12 months in men. The post hoc statistical power (%) of FN BMD loss within 6 months in the women group was 93.82%, and for the FN BMD loss within 12 months in the women and men group, it was 86.88% and 55.56% respectively. Sex-specific responses of SG revealed a more significant bone loss in men than in women at TH (a loss of BMD of 8.62% vs 5.44% across 12 months). The difference in FN BMD changes between women and men did not reach statistical significance ( Table 3 ). However, LS BMD showed no changes at 6 months and 12 months postoperatively in both sexes (p > 0.05).

The changes in measured bone turnover markers were shown in Figure 2 . Compared with the baseline, all measured bone turnover markers increased for 12 months. In women, the P1NP and N-MID levels (56.54 ng/mL and 14.60 ng/mL, respectively) increased persistently to peak values at 12 months postoperatively (68.82 ng/mL and 19.63 ng/mL, respectively; both P < 0.001). The β-CTX levels continued to increase and reached its peak at 6 months (356.90 vs. 839.31 pg/mL; P < 0.05), remaining at a high level until 12 months (356.90 vs. 646.42 pg/mL; P < 0.001). Similarly in men, the P1NP and N-MID levels (55.79 ng/mL and 15.75 ng/mL, respectively) increased persistently to peak values at 12 months postoperatively (97.01 ng/mL and 37.88 ng/mL, respectively; both P < 0.001). The β-CTX levels continuously increased, reaching a peak at 6 months (493.67 vs. 1329.63 pg/mL; P < 0.001) and remaining at a high level until 12 months (493.67 vs. 1072.26 pg/mL, P < 0.001). The GLMMs confirmed higher levels of P1NP and β-CTX in men versus women at both 6 months and 12 months. The difference in N-MID level changes between women and men did not reach statistical significance ( Table 3 ).

To detect potential body composition variables related to the BMD loss, correlation analysis and subsequent stepwise multiple linear regression analysis were performed in both sexes for 6 months and 12 months after SG separately. During the first 6 months post-SG, the FN BMD loss was positively associated with the percent reduction in eFFMI after adjusting age and %EWL in women (adjusted β = 0.77, P = 0.004, adjusted R^2 = 0.54) ( Supplementary Table 3 ), while the FN BMD loss was positively associated with the percent reduction in SFA in men (r = 0.93, P = 0.007) ( Supplementary Table 1 ). For 12 months post-SG, the FN BMD loss was negatively associated with the percent reduction in VFA in both women and men. The significance remained after adjusting age and percentage change in waist circumference in women (adjusted β = -0.58, P = 0.027, adjusted R^2 = 0.42) and adjusting age as well as percentage change in total triglycerides in men (adjusted β = -0.68, P = 0.032, adjusted R^2 = 0.39) ( Supplementary Table 3 ). For TH, TH BMD loss was positively associated with a percent reduction in waist circumference (r = 0.45, P = 0.028) in women ( Supplementary Table 2 ). Sex-specific factors were found in different bone sites. Furthermore, changes in various body compositions are associated with FN BMD loss at different time points. This indicated that changes in body composition might play roles in BMD loss in a dynamic and nonlinear fashion.