Gastric Bypass But Not Sleeve Gastrectomy Increases Risk of Major Osteoporotic Fracture: French Population‐Based Cohort Study

ABSTRACT

The objective of this work was to investigate the risk of major osteoporotic fracture (MOF; hip, proximal humerus, wrist and distal forearm, and clinical spine) in bariatric surgery patients versus matched controls. Bariatric surgery is associated with an increase in fracture risk. However, it remains unclear whether the same degree of fracture risk is associated with sleeve gastrectomy, which has recently surpassed gastric bypass. Records from the French National Inpatient database were used from 2008 to 2018. Bariatric surgery patients, aged 40 to 65 years, with BMI ≥40 kg/m², hospitalized between January 1, 2010 and December 31, 2014, were matched to one control (1:1) by age, sex, Charlson comorbidity index, year of inclusion, and class of obesity (40 to 49.9 kg/m² versus ≥50 kg/m²). We performed a Cox regression analysis to assess the association between the risk of any MOF and, respectively, (i) bariatric surgery (yes/no) and (ii) type of surgical procedure (gastric bypass, gastric banding, vertical banded gastroplasty, and sleeve gastrectomy) versus no surgery. A total of 81,984 patients were included in the study (40,992 in the bariatric surgery group, and 40,992 matched controls). There were 585 MOFs in the surgical group (2.30 cases per 1000 patient‐year [PY]) and 416 MOFs in the matched controls (1.93 cases per 1000 PY). The risk of MOF was significantly higher in the surgical group (hazard ratio [HR] 1.22; 95% CI, 1.08–1.39). We observed an increase in risk of MOF for gastric bypass only (HR 1.70; 95% CI, 1.46–1.98) compared with the matched controls. In patients aged 40 to 65 years, gastric bypass but not sleeve gastrectomy or the other procedures increased risk of major osteoporotic fractures. © 2020 American Society for Bone and Mineral Research.

BARIATRIC SURGERY, FRACTURES, GASTRIC BYPASS, OBESITY, OSTEOPOROSIS, SLEEVE GASTRECTOMY

Introduction

The prevalence of overweight and obesity has been continuously on the rise throughout the world. Obesity is a major public health issue worldwide.¹ Morbid obesity (BMI ≥40 kg/m²) has followed the same dramatically increasing trend. Unfortunately, various efforts aimed at weight loss, including exercise, diet intervention, or other nonsurgical procedures, are ineffective in the long term in reducing weight and maintaining weight loss in morbidly obese individuals.² Surgical treatment of morbid obesity has been increasingly used and has many benefits beyond weight loss, including dramatic remission of type 2 diabetes, and improvement in blood pressure and dyslipidemia.³ Unfortunately, bariatric surgery is also associated with complications. Medical complications include gastroesophageal reflux disease, malnutrition, and metabolic complications deriving from vitamin and mineral malabsorption.

More recently, it has been increasingly acknowledged that bariatric surgery adversely affects skeletal health. The extent of high‐turnover bone loss is much greater than what would be expected in the absence of a severe skeletal insult.⁴ Patients also experience a significant deterioration in bone microarchitecture, as assessed by high‐resolution peripheral quantitative computed tomography.^{5,  6} In their seminal study, in a retrospective cohort, Lalmohamed and colleagues⁷ found no significant increase in fracture risk after bariatric surgery, but there is now a growing body of evidence that suggests an association between bariatric surgery and higher fracture risk.^{8,  9,  10,  11,  12,  13,  14}

However, there are still many unknowns in this field. It is yet unclear whether the increase in fracture risk in older men and women is associated with bone fragility because studies have mainly been conducted in younger populations (average age ~43 years) than those at risk of osteoporotic fractures.^{8,  9,  10,  11,  12,  13} It also remains unclear whether the same degree of fracture risk is associated with sleeve gastrectomy, which has recently surpassed gastric bypass as the most popular form of bariatric surgery.^{15,  16,  17} Considering these knowledge gaps, a larger population‐based study, including older patients and a large number of patients undergoing sleeve gastrectomy was needed. We conducted this nationwide cohort study with the hypothesis that bariatric surgery would increase the risk of major osteoporotic fracture‐related hospitalization (MOF) in patients aged 40 to 65 years. Our main objective was to assess the relationship between the risk of MOF and the fact of having undergone bariatric surgery (yes/no) in morbidly obese patients. Our secondary objectives were to determine the relationship between the risk of MOF among different surgical types and fractures types.

Patients and Methods

Data sources

For the purposes of our study, we collected data from the French National Inpatient database, spanning the period between 2008 and 2018. The data contain extensive information on inpatient stays in all French public and private hospitals (270 million inpatient stays for the period in question). They include diagnoses encoded according to the International Classification of Diseases, 10th revision (ICD‐10), medical procedures according to the French classification of medical procedures (CCAM), and demographic and administrative data. The collection of the data was approved by the French data protection authority (Commission Nationale de l'Informatique et des Libertés [CNIL] authorization number 2049035), and the data were anonymized before being analyzed.

To ensure the quality of the data, 153 tests¹⁸ are performed routinely when information on inpatient stays is sent to the French public health insurance agency. These tests include checks on the chronology of the inpatient stays, the format (missing, incorrect or imprecise values) of the demographic characteristics (sex, age, date and mode of entry, date and mode of discharge), the format of procedure and diagnostic codes, and the consistency between procedure codes, diagnostic codes, length of stay, age, and sex.

Study design

We proposed a retrospective cohort with two matched groups of subjects included from January 1, 2010 to December 31, 2014: a group of obese patients who had undergone bariatric surgery, and a group of non‐operated obese patients. All patients were followed for up to 8 years.

Participants

In France, as is the case in other countries, current medical guidelines require that patients take part in a medical management program before undergoing surgery. Eligibility was also based on body mass index (BMI) ≥40 or ≥35 kg/m² with one or more comorbidity. Surgery before 18 and after 60 years was considered only on a case‐by‐case basis.

We began by forming a cohort of subjects aged 40 to 65 years, with a BMI of at least 40 kg/m². The main inclusion criterion was the existence of ICD‐10 codes for morbid obesity (BMI 40 to 49.9 kg/m²) or super obesity (BMI ≥50 kg/m²) during inpatient stays (codes listed in Supporting Table 1).

We excluded patients with a history of (i) bone disease other than osteoporosis (eg, Paget disease, multiple myeloma), (ii) any cancer, or (iii) bariatric surgery reported from January 1, 2008 until inclusion (the corresponding ICD‐10, CCAM, and French diagnosis‐related group (DRG) codes are listed in Supporting Tables 2 and 3). For each patient, the first inpatient stay from January 1, 2010 to December 31, 2014 was considered for inclusion: if the patient had undergone surgery in the 6 months following this first inpatient stay, they were assigned to the surgery group and included from this moment. Moreover, subjects were excluded if they died or fractured during the first inpatient stay (Supporting Fig. 1).

Matching

Each bariatric surgery patient was matched to one control (1:1) by age, sex, class of obesity (BMI 40 to 49.9 kg/m² and BMI ≥50 kg/m²), year of inclusion, and Charlson comorbidity index (CCI), which is a prognostic index developed to predict 1‐year mortality among patients admitted to the medical department of an acute care hospital.¹⁹ The CCI was computed using ICD‐10 codes, as proposed by Quan and colleagues.²⁰

Exposures of interest

The main exposure of interest was the fact of having undergone (or not) a bariatric surgery procedure at baseline.

Furthermore, among the morbidly obese subjects, individuals who had undergone bariatric surgery were identified using the CCAM codes for, respectively, gastric bypass, gastric banding, vertical banded gastroplasty (VBG), and sleeve gastrectomy. The corresponding codes are listed in Supporting Table 2. Because of the low number of biliopancreatic diversion performed from January 1, 2010 to December 31, 2014, we did not include individuals who had undergone this type of surgical procedure (identified using the CCAM code for biliopancreatic diversion).

Main outcomes

The main outcome was the first diagnosis of any MOF‐related hospitalization (ie, hip, proximal humerus, wrist and distal forearm, and clinical spine) from the time of inclusion (the corresponding ICD‐10 codes are listed in Supporting Table 4). In addition, fracture site‐specific analyses (hip, proximal humerus, wrist and distal forearm, and clinical spine) were also considered as outcomes of interest.

Other variables of interest

In addition to the variables involved in the matching, the following variables were collected prior to the index date (ie, from January 1, 2008 to the date of inclusion): diabetes mellitus, cardiovascular diseases, chronic pulmonary diseases, and history of any osteoporotic fracture (craniofacial, cervical spine, hand, finger, foot, and toe fractures were excluded).

Statistical analysis

A preliminary descriptive analysis was performed to establish the patients’ baseline characteristics. Categorical variables were expressed as frequency and absolute number. Quantitative variables were expressed as mean ± standard deviation (SD).

To assess the association between the time to occurrence of at least one MOF in the two groups of interest, we used a Cox proportional hazard model on matched individuals, adjusted for the variable “history of any osteoporotic fracture.” Censoring criteria were death, loss/end of follow‐up, and bariatric surgery in the control group. Matching was performed based on the variables described above in the “matching” section. The proportional hazards (PH) assumption was assessed using the Schoenfeld residuals test and the corresponding test for trend. As the PH assumption was violated (p < .001), a piecewise Cox model (with a time‐dependent effect) was used and demonstrated a good fit. The interaction considered in this model involved the bariatric surgery groups (gastric bypass, sleeve gastrectomy, adjustable gastric banding, VBG, and no surgery) and time (the latter considered as eight categories of 1 year).

A similar Cox regression model was performed to assess the association between the occurrence of at least one MOF and each type of procedure (gastric bypass, gastric banding, VBG, sleeve gastrectomy, and no surgery). This analysis was performed using an explanatory variable with five categories, the control group being used as the reference group. Similarly, another analysis was conducted to assess the relationship between bariatric surgery and the risk of site‐specific fractures. In this last analysis, the patient is censored at the date of occurrence of the first fracture even if this fracture is not the fracture of interest. All patients included in the study were analyzed in each of the models described in the two paragraphs above.

Tests of statistical significance were performed at the two‐tailed alpha level of 0.05. The analysis was carried out using the R Statistical Software (version 3.3.1; R Foundation for Statistical Computing, Vienna, Austria) and the “survival” package.^{21,  22}

Results

Baseline characteristics

Table 1 shows the baseline characteristics of the bariatric surgery patients and matched controls. We identified 40,992 patients who had undergone bariatric surgery (mean age 49.1 years; 78.4% female), and a total of 40,992 matched controls (mean age 49.1 years; 78.4% female). Sleeve gastrectomy was the most frequent surgical technique for bariatric surgery (18,635; 45.5%), followed by gastric bypass (14,532; 35.4%). Matched controls were more likely to have a history of any osteoporotic fracture in the years before inclusion (0.9% versus 0.6%). Total duration of follow‐up was 469,814 person‐years. The average follow‐up duration was 5.73 years for all participants with a mean of 6.19 years in the surgical group (gastric bypass 6.23 years; gastric banding 6.79 years; VGB 7.58 years; sleeve gastrectomy 5.79 years) and 5.26 years in the control group. During the follow‐up, 1180 patients died (1.4%): 826 in the control group (no surgery) and 354 in the bariatric group. Moreover, during the follow‐up, 21.7% of matched controls had undergone surgery after inclusion and had been censored.

Baseline Characteristics of Groups

Characteristic	Bariatric (n = 40,992)	Obese (n = 40,992)	Gastric bypass (n = 14,532)	Sleeve gastrectomy (n = 18,635)	Gastric banding (n = 5178)	Vertical banded gastroplasty (n = 2647)
Women	32,140 (78.4)	32,140 (78.4)	11,796 (81.2)	14,253 (76.5)	4121 (79.6)	1970 (74.4)
Age (years), mean ± SD	49.1 (6.6)	49.1 (6.6)	49.1 (6.5)	49.1 (6.7)	48.4 (6.5)	49.7 (6.7)
Age (classes)
40–49 years	25,271 (61.6)	25,271 (61.6)	8859 (61.0)	11,516 (61.8)	3368 (65.0)	1528 (57.7)
50–59 years	13,021 (31.8)	13,021 (31.8)	4783 (32.9)	5822 (31.2)	1507 (29.1)	909 (34.3)
60–65 years	2700 (6.6)	2700 (6.6)	890 (6.1)	1297 (7.0)	303 (5.9)	210 (7.9)
BMI ≥ 50 kg/m2	5178 (12.6)	5178 (12.6)	1761 (12.2)	2474 (13.3)	459 (8.9)	478 (18.1)
Diabetes	5896 (14.4)	5996 (14.6)	2468 (17.0)	2389 (12.8)	553 (10.7)	486 (18.4)
COPD	1762 (4.3)	1541 (3.8)	625 (4.3)	784 (4.2)	198 (3.8)	155 (5.9)
CVD	411 (1.0)	1141 (2.8)	137 (0.9)	180 (1.0)	54 (1.0)	40 (1.5)
Charlson comorbidity index
0	31,846 (77.7)	31,846 (77.7)	10,952 (75.4)	14,646 (78.6)	4312 (83.3)	1936 (73.1)
1	7581 (18.5)	7581 (18.5)	2936 (20.2)	3352 (18.0)	734 (14.2)	559 (21.1)
2	1226 (3.0)	1226 (3.0)	500 (3.4)	506 (2.7)	105 (2.0)	115 (4.3)
≥3	339 (0.8)	339 (0.8)	144 (1.0)	131 (0.7)	27 (0.5)	37 (1.5)
History of any osteoporotic fracture (before index date)	234 (0.6)	372 (0.9)	76 (0.5)	121 (0.6)	31 (0.6)	6 (0.2)
Year of inclusion
2010	6755 (16.5)	6755 (16.5)	2694 (18.5)	459 (2.5)	1531 (29.6)	2071 (78.2)
2011	6950 (17.0)	6950 (17.0)	2489 (17.1)	2999 (16.1)	1195 (23.1)	267 (10.1)
2012	8489 (20.7)	8489 (20.7)	2936 (20.2)	4267 (22.9)	1121 (21.6)	165 (6.2)
2013	9201 (22.4)	9201 (22.4)	3225 (22.2)	5124 (27.5)	763 (14.7)	89 (3.4)
2014	9597 (23.4)	9597 (23.4)	3188 (21.9)	5786 (31.0)	568 (11.0)	55 (2.1)

Values are n (%) unless stated otherwise.

BMI = body mass index; COPD = chronic obstructive pulmonary disease; CVD = cardiovascular disease.

Comparison of fracture risk between bariatric surgery patients and matched controls

Table 2 shows the overall risk of MOF in bariatric surgery patients compared with matched controls, by fracture type. At the end of the study period, there were a total of 585 MOFs in the surgical group (2.30 cases per 1000 patient‐year [PY]) and 416 MOFs in the matched controls (1.93 cases per 1000 PY). We observed an increase in overall risk for any MOF with adjusted hazard ratio (HR) of 1.22 and 95% CI of 1.08 to 1.39.

HR of Major Osteoporotic Fracture Between Bariatric Surgery and No Surgery Groups, by Fracture Site

Outcome	Group	Events (n)	HR (95% CI)5	p value
Major osteoporotic fractures	Bariatric group	585	1.22 (1.08–1.39)	.002
	No surgery	416	Reference
Wrist and forearm	Bariatric group	303	1.61 (1.33–1.95)	<.001
	No surgery	161	Reference
Humerus	Bariatric group	113	0.74 (0.57–0.95)	.02
	No surgery	130	Reference
Clinical spine	Bariatric group	78	1.07 (0.77–1.45)	.67
	No surgery	64	Reference
Hip	Bariatric group	91	1.38 (0.99–1.91)	.06
	No surgery	61	Reference

n = 81 984; total duration of follow up was 469,814 person‐years: 215,936 person‐years in the control group and 253,878 person‐years in the bariatric group.

Cox regression analysis including a stratum indicator variable and an adjustment on the variable “history of any osteoporotic fracture.”

When fracture site‐specific analyses were performed, patients in the surgical group were found to have a higher risk of fracture at the distal forearm and wrist (HR 1.61; 95% CI, 1.33–1.95) and a lower risk of fracture at the proximal humerus (HR 0.74; 95% CI, 0.57–0.95).

Comparison of fracture risk by type of bariatric procedure

When fracture risk was analyzed by type of procedure, we observed a higher risk of MOF for gastric bypass only (HR 1.70; 95% CI, 1.46–1.98) compared with the matched controls (Table 3). Moreover, we did not observe a higher risk of MOF either for sleeve gastrectomy (HR 0.95; 95% CI, 0.79–1.14) or for adjustable gastric banding (HR 0.95; 95% CI, 0.72–1.25) compared with the matched controls (Table 3).

HR of Major Osteoporotic Fracture Between Types of Bariatric Procedures and No Surgery

Outcome	Group	Person‐years	Events (n)	HR (95% CI)7	p
Major osteoporotic fractures	Roux‐en‐Y gastric bypass	90,628	292	1.70 (1.46–1.98)	<.001
	Sleeve gastrectomy	107,967	183	0.95 (0.79–1.14)	.612
	Adjustable gastric banding	35,202	64	0.95 (0.72–1.25)	.717
	Vertical banded gastroplasty	20,080	46	0.95 (0.68–1.31)	.740
	No surgery	215,936	416	Reference

n = 81,984 refers to the bariatric surgery patients and matched controls.

Cox regression analysis including a stratum indicator variable and an adjustment on the variable “history of any osteoporotic fracture.”

Figure 1 shows the nonadjusted fracture‐free survival rates (ie, no MOFs) for the types of bariatric procedures and matched controls. Although the rates for the other procedures (sleeve gastrectomy, adjustable gastric banding, and vertical banded gastroplasty) and obese groups are comparable, they decreased more rapidly in the gastric bypass group.

Nonadjusted fracture‐free survival rate (no MOFs) between types of bariatric procedures and matched controls. MOF = major osteoporotic fracture.

Comparison of fracture risk by subgroups of bariatric procedure and matched controls

Table 4 shows the risk of MOF in the sleeve gastrectomy, adjustable gastric banding, vertical banded gastroplasty, and gastric bypass groups compared with matched controls, by fracture type. Patients in the gastric bypass group had a higher risk of hip fracture (HR 2.24; 95% CI, 1.53–3.26) and distal forearm and wrist fractures (HR 2.40; 95% CI, 1.93–2.99). Patients in the sleeve gastrectomy group had a lower risk of proximal humerus fracture (HR 0.65; 95% CI, 0.45–0.94).

HR of Major Osteoporotic Fracture Between Gastric Bypass, Sleeve Gastrectomy, Gastric Banding, VBG, and No Surgery Groups, by Fracture Site

Outcome	Group	Events (n)	HR (95% CI)9	p value
Wrist and forearm	Gastric bypass	167	2.40 (1.93–2.99)	<.001
	Sleeve gastrectomy	88	1.20 (0.92–1.57)	.17
	Gastric banding	27	1.03 (0.68–1.55)	.90
	VBG	21	1.09 (0.67–1.76)	.73
	No surgery	161	Reference
Humerus	Gastric bypass	47	0.88 (0.62–1.23)	.44
	Sleeve gastrectomy	40	0.65 (0.45–0.94)	.02
	Gastric banding	17	0.76 (0.45–1.28)	.30
	VBG	9	0.57 (0.28–1.16)	.12
	No surgery	130	Reference
Clinical spine	Gastric bypass	25	1.02 (0.64–1.63)	.93
	Sleeve gastrectomy	31	1.01 (0.65–1.58)	.95
	Gastric banding	12	1.13 (0.60–2.13)	.70
	VBG	10	1.44 (0.70–2.98)	.32
	No surgery	64	Reference
Hip	Gastric bypass	53	2.24 (1.53–3.26)	<.001
	Sleeve gastrectomy	24	0.65 (0.45–0.94)	.60
	Gastric banding	8	0.95 (0.44–2.04)	.90
	VBG	6	0.88 (0.36–2.16)	.79
	No surgery	61	Reference

n = 81,984 refers to the bariatric surgery patients and matched controls.

VBG = vertical banded gastroplasty.

Cox regression analysis including a stratum indicator variable and an adjustment on the variable “history of any osteoporotic fracture.”

Comparison of fracture risk between subgroups by period

The HRs of MOF in the sleeve gastrectomy, adjustable gastric banding, vertical banded gastroplasty, and gastric bypass groups relative to the risk in matched controls (without surgery) was computed using the time‐dependent Cox model (Fig. 2). The risk of MOF after gastric bypass varied over time. In contrast, this effect was not found for other procedures (Fig. 2): there were no significant differences in MOF risk between the gastric bypass group and matched controls in the first 3 years after surgery, but there was a progressive increase in fracture risk in subsequent years.

Hazard ratios representing adjusted risk of major osteoporotic fracture over time for each group.

Discussion

In this large‐scale study of bariatric surgery patients in France—mainly women in their late 40s—we showed that severely obese patients who had undergone surgery and were followed for up to 8 years are at higher risk of MOF compared with severely obese controls. Furthermore, only gastric bypass was clearly associated with a higher risk of MOF. Our study also suggests that fracture risk is site specific, with a higher risk of distal forearm and wrist fractures, which is typical of first osteoporotic fracture in young menopausal women, and a lower risk of proximal humerus fracture, which is typical of fracture in obese patients.

Several epidemiologic studies to date have examined fracture risk after bariatric surgery^{7,  8,  9,  10,  11,  12,  13,  14} (Supporting Table 5). In a study conducted by Lalmohamed and colleagues,⁷ the authors found no increase in fracture risk (osteoporotic and non‐osteoporotic) in patients who had undergone bariatric surgery compared with controls (matched for age, sex, practice, year, and BMI), but the follow‐up time was relatively short (median time 2.2 years for bariatric surgery patients). No association was detected between bariatric surgery and fractures in another study performed in the UK. However, this study included a large number of patients undergoing gastric banding procedure (47.1%).¹³ In a study conducted by Lu and colleagues,⁸ the authors reported a significant 1.21‐fold (range, 1.02–1.43) increase in fracture risk (osteoporotic and non‐osteoporotic) in the surgical group compared with the obese control group (matched for age, sex, CCI, diabetes, hypertension, hyperlipidemia, and year obesity was diagnosed). However, they found no evidence of a higher risk of osteoporotic fracture (HR 1.05; 95% CI, 0.77–1.43). When stratified by surgical procedures, a higher risk of fracture (osteoporotic and non‐osteoporotic) was found for malabsorptive procedures alone (HR 1.47; 95% CI, 1.01–2.15) in comparison with controls,⁸ which is in line with our findings. In a study conducted by Rousseau and colleagues,⁹ bariatric patients followed for a mean of 4.4 years after surgery were found to be more prone to fracture compared with the obese group (relative risk 1.38; 95% CI, 1.23–1.55). After surgery, only biliopancreatic diversion was clearly associated with a higher risk of fracture (adjusted relative risk 1.60; 95% CI, 1.25–2.03) compared with the non‐obese group, whereas no comparison was available with the obese control group.⁹ Using claims data from a US commercial health plan, Yu and colleagues¹⁰ compared the risk of nonvertebral osteoporotic fracture (including humerus, wrist, hip, and pelvis fractures) in obese adults after gastric bypass and gastric banding procedures and found a higher risk of nonvertebral osteoporotic fracture (HR 1.43; 95% CI, 1.13–1.81) in gastric bypass patients compared with gastric banding patients. Two other studies were powered to evaluate gastric bypass–specific fracture risk in large population data sets.^{11,  14} Using Medicare, Yu and colleagues¹⁴ found that gastric bypass was associated with an increased risk of nonvertebral fractures, including hip, wrist, and pelvis fractures, in comparison with gastric banding. Moreover, older adults had similar gastric bypass–associated increases in fracture risk as younger adults. In a Swedish study performed by Axelsson and colleagues¹¹ gastric bypass in comparison with (obese) controls was associated with increased risk of any fracture, in patients with and without diabetes. Moreover, larger weight loss or poor calcium and vitamin D supplementation after surgery were not associated with increased fracture risk.¹¹ Fashandi and colleagues²³ also found that patients who had bariatric surgery, mainly gastric bypass (~80% of bariatric surgery procedures), are at increased risk of any fracture compared to a propensity‐matched control group (6.4% versus 2.7%; p = .0001).

Regarding fracture site‐specific analyses, no comparisons between the bariatric surgery group and the obese group were reported in the study by Rousseau and colleagues⁹ In the study conducted by Lu and colleagues,⁸ only unusual fracture sites—including the clavicle, scapula, sternum, and feet and toes—reached statistical significance. Using two different US databases, Yu and colleagues^{10,  14} found that gastric bypass patients compared to gastric banding had a higher risk of fracture at the hip, and wrist, which is in line with our findings. Moreover, the Swedish study also support our findings of more hip and upper‐extremity fractures after gastric bypass, although they found a paradoxically reduced risk of lower‐leg fracture.¹¹ Rousseau and colleagues⁹ also found a reduced risk of lower‐leg fracture after bariatric surgery. This fracture site was associated with obesity in the study by Rousseau and colleagues⁹ and fracture risk at this site may thus be reduced after weight loss. In our study, only sleeve gastrectomy was associated with a lower risk of proximal humerus fractures. Interestingly, a higher risk of proximal humeral fractures has been associated with obesity in some studies,^{24,  25} and fracture risk at this site may thus be reduced after weight loss. However, the protective effect of sleeve gastrectomy on proximal humerus fracture risk needs further validation.

It is interesting to note that fracture risk in our study starts to rise after 3 years postoperation, which is similar to what was reported in other studies.^{10,  11,  14} In the study by Yu and colleagues,¹⁰ nonvertebral fracture risk associated with gastric bypass manifested >2 years after surgery and increased in subsequent years, with the highest risk in the fifth year after surgery (HR 3.91; 95% CI, 1.58–9.64). This increased fracture risk, gradually increasing in time, has also been demonstrated in two other studies.^{11,  14} In our population, this means that fracture risk starts to increase when the mean age of the population is about 52 years. Given that 80% of our population is women, this would coincide with the entry in menopause for most women. Moreover, in the study performed by Rousseau and colleagues,⁹ they showed that there seemed to be a second peak of fracture at around year 11, which corresponds to the age of menopause for most women in this study. Furthermore, the study by Schafer and colleagues⁵ has shown that gastric bypass has a greater impact on bone in postmenopausal women.

Although the mechanism underlying the increase in bone remodeling and fracture risk after bariatric surgery is not fully understood, many factors (eg, estrogen, adipokines, weight loss, deficiencies in nutrients) seem to be involved. It seems plausible that surgeries that include a malabsorptive component, such as gastric bypass, lead to greater bone loss than do other procedures, because they induce greater weight loss and are more likely to cause deficiencies in nutrients that are important for bone health, including calcium, vitamin D, and protein.^{26,  27,  28} However, in the study performed by Axelsson and colleagues,¹¹ larger weight loss or poor calcium and vitamin D supplementation after surgery were not associated with increased fracture risk. In line with this, skeletal changes after gastric bypass (high‐turnover bone remodeling markers and bone microarchitectural parameters declines) persist for up to 5 years after weight loss has plateaued and weight stabilizes.^{28,  29} On the other hand, an increased risk of fall injury following gastric bypass surgery has been recently found by Axelsson and colleagues.¹¹ Interestingly, observational evidence suggests that patients’ physical functioning, including 6‐min walk test (6MWT), Timed Up‐and‐Go (TUG) test, and Short Physical Performance Battery (SPPB), improves following bariatric surgery,³⁰ although a drastic muscle mass loss also occurs, particularly in the first 6 postoperative months.³¹ Comprehensive reviews have also found significant postoperative increases in physical activity, as assessed by self‐reported and objective measures including walking performance.³² We can hypothesize that improved functioning lead to greater physical activity, which is likely to result in increased exposure and therefore greater risk of falls injury such as fractures at the wrist/forearm in young postmenopausal women.

Changes in body composition and bone marrow adiposity may also contribute in the deleterious bone effects of bariatric surgery.^{33,  34,  35} Indeed, bariatric surgery results in loss of muscle mass particularly during the first 6 months after surgery and after that changes in muscle mass are highly variable from muscle mass gain to persistent loss of muscle mass depending on the nutritional status and the level of physical activity.³⁶ Regarding the link between muscle metabolism and bone health, it seems possible that loss of muscle mass exacerbates the deleterious effect of bariatric surgery on bone health, whereas muscle mass gain could counteract the loss of bone mass and quality.³⁷ Findings from few studies suggest that bone marrow adiposity could also contribute to negative skeletal effects of bariatric surgery.^{15,  38,  39,  40} In a pilot study involving morbidly obese diabetic (n = 6) and nondiabetic women (n = 5) undergoing gastric bypass surgery, Schafer and colleagues³⁸ examined the effects of gastric bypass on vertebral bone marrow fat (BMF) content using magnetic resonance spectroscopy (MRS). Six months postoperatively, in those without diabetes, BMF was maintained on average after gastric bypass (+0.9%), despite dramatic declines in overall fat mass. In those with diabetes, gastric bypass significantly reduced BMF content (−7.5%; p = .05).³⁸ With an extended cohort of 30 women (14 women with diabetes and 16 without), the same team examined vertebral BMF content and BMD changes over a period of 6 months post–gastric bypass.³⁹ In women with diabetes, gastric bypass significantly reduced BMF content (−6.5%; p = .05), whereas in the nondiabetic women, BMF content was stable (+1.8%; p = .29).³⁹ However, the skeletal effects of bariatric surgery are complex and, in another bariatric study, the authors did not report any changes in BMF content in subgroups of participants who had had gastric bypass, although they did find an association between sleeve gastrectomy and an increase in marrow fat.¹⁵ Recently, Blom‐Høgestøl and colleagues⁴⁰ found that bone marrow adipose tissue (BMAT) fraction, assessed by bone marrow biopsies decreased, with a mean percent of 10.7% at 12 months after gastric bypass, but without any statistical difference in BMAT reduction in patients with and without diabetes mellitus. Then, results are inconsistent and further studies are needed to better understand the impact of bariatric surgery on BMF depending on diabetic status, gender, changes in total body fat, and type of bariatric surgery (eg, gastric bypass, sleeve gastrectomy).

The major strengths of this study include the large number of subjects in the study cohort and the number of MOFs, which increased statistical power and enabled an examination of each type of MOF and surgery procedure. Furthermore, we are confident that fractures and study groups were properly identified, because we used an algorithm for fracture identification, intervention codes for bariatric surgery, and stringent criteria to identify obese people.

The study does, however, have limitations. Surgical subgroups were not necessarily comparable in terms of BMI and comorbidities, which could partly explain differences in fracture risk among bariatric procedures. However, our groups were matched for body mass index categories and for CCI. We decided to include all patients who had undergone bariatric surgery to ensure that this population was representative of the bariatric population. Previous fracture history or osteoporosis risk factors are not contraindications for surgery, and previous fracture history is a strong risk factor for future fractures. As such, excluding patients with a previous fracture history would exclude patients who are at high risk of fracture after bariatric surgery. Furthermore, no information on vertebral fractures without clinical symptoms and inpatient stay was captured, and this might have resulted in an underestimate. In addition, items of the CCI and history of fracture prior to the time of inclusion were sought only for a relatively short period of time (between January 1, 2008 and the date of inclusion), which may also present a limitation in the quality of the data used in this study. Moreover, we acknowledge that the method of using all available data prior to index date leads to variability in the assessment time among the included patients. Nevertheless, it seems reasonable to think that such a bias would be nondifferential (between the exposures of interest) because we used the year as a matching criteria. Another limitation is that the level of trauma and several important confounders that may affect bone health or risk of falls—including drugs, physical activity, weight loss, smoking, alcohol consumption, and use of vitamin D and calcium supplements—were not assessed. In addition, we had no information on BMD. We also speculate that the use of bisphosphonates and hormone replacement therapy may have biased the results in a very small proportion of the study group. We cannot exclude the possibility of confounding by contraindication or indication in this study. We had no information on whether control patients were considered for bariatric surgery and then did not undergo an operation because of contraindication. Indeed, non‐operated obese participants were included because they were hospitalized. We acknowledge that this may introduce another bias because this non‐operated obese population is probably sicker than the general obese population. We also acknowledge that, in identifying controls, there are likely to be inaccuracies in coding for obesity or BMI categories on the basis of a single ICD‐10 code for obesity. Moreover, regarding the surgery types analysis, because the matching was carried out for surgery/no surgery groups, some strata do not contain all surgery types. However, it seems to be unlikely that this situation biased the results because the surgery types appear to be quite equally distributed according to covariates. Finally, our follow‐up time was relatively short, which reduced the power to exclude an increase in fracture risk for other procedures.

Conclusion

In conclusion, severely obese patients who have undergone bariatric surgery are more prone to osteoporotic fracture than are matched controls. Moreover, the risk of fracture was associated with gastric bypass in the present study. Finally, the increase in fracture risk observed with gastric bypass should be considered to determine the type of bariatric procedure, especially in patients at higher risk of osteoporosis.

Disclosures

All authors state that they have no conflicts of interest.

Acknowledgments

The authors are employed by their university and/or their hospital. These funding organizations did not suggest the subject of this study, and did not have access to the results before publication. This study received no external funding.

Authors' roles: JP had the idea for the study. JP and GF designed the study. NM, EB, and GF did the statistical analyses. JP, EL, EB, BC, and GF interpreted the findings. JP and GF wrote the first draft of the manuscript and revised subsequent versions. The other authors provided input, expertise, and critical review of the manuscript. All authors read and approved the final version of the manuscript. JP and GF are the guarantors.

Author notes

The peer review history for this article is available at https://publons.com/publon/10.1002/jbmr.4012.