Obesity remains a growing health concern and is associated with many comorbid medical conditions including hypertension, diabetes, dyslipidemia, coronary disease, stroke, and sleep apnea. In America, more than two-thirds (68.8%) of the adult population is considered overweight (BMI > 25 kg/m2) and more than one-third (35.7%) is considered obese (BMI >30 kg/m2) [1, 2]. Bariatric surgery is increasingly being used to combat the obesity epidemic and is currently used to treat patients with a BMI greater than 40 kg/m2 or those with a BMI greater than 35 kg/m2 who have obesity-related comorbid disease. Approximately 179,000 weight-loss surgeries were performed in 2013 in the US alone, most of which were Roux-en-Y gastric bypass (RYBG, 34.2%) or sleeve gastrectomy (SG, 42.1%) procedures, with fewer adjustable gastric banding (14%) and occasional biliopancreatic diversion (BPD) with duodenal switch (1%) procedures [3]. Bariatric surgery has been shown to have outstanding results with both durable weight loss and sustained reduction of obesity-related comorbid disease [4,5,6].

Although bariatric surgery has many positive effects on patient health, there is some evidence of associated health risks, including an increased risk for bone disease. Several studies have shown that bariatric surgery leads to increased bone resorption markers [7, 8] and reduced bone mineral density [9]. These detrimental effects on bone health are multifactorial and likely involve nutritional deficiencies, changes in mechanical loading and alterations in various gastrointestinal and fat-associated hormonal factors [3, 8, 10,11,12]. Given these bone changes, several studies have been undertaken to assess the risk of bone fracture after bariatric surgery [13,14,15,16]. The results of these studies are conflicting; however, the true effect of bariatric surgery on skeletal fragility remains unknown.

The goal of this study was to determine the long-term incidence of bone fracture after bariatric surgery and to identify risk factors for fracture. We additionally aimed to compare the risk of fracture in a group of individuals who underwent bariatric surgery to that of a group of propensity-matched morbidly obese patients who did not undergo weight-loss surgery. We hypothesized that bariatric surgery would independently increase a patient’s long-term risk of fracture.

Materials and methods

The University of Virginia Institutional Review Board approved this study (IRB Protocol #17132). All adults undergoing bariatric surgery at a single institution between 1985 and 2015 were reviewed from a previously validated institutional database [4]. This database included demographic data as well as preoperative comorbidities [diabetes mellitus, gastroesophageal reflux disease, obstructive sleep apnea (OSA), hypertension, osteoarthritis, pulmonary disease, and cardiac disease], operative details, and postoperative complications. Additionally, the institutional Clinical Data Repository (CDR) captures all patient visits recorded in our electronic medical record and was used to identify patients within our bariatric database who were evaluated or treated for a bone fracture. Patients were stratified by fracture status and univariate analyses characterized baseline and operative characteristics. The Mann–Whitney U test was used for independent continuous variables and the Wilcoxon signed-rank sum test for paired ordinal data. The χ2 test was used for independent categorical variables. Multivariate logistic regression was utilized to identify predictors of long-term bone fracture. A priori, the decision was made to include all clinically reasonable preoperative variables and no selection methods were used.

To identify an appropriate control group, the CDR was used to find a matched cohort of morbidly obese patients not undergoing bariatric surgery. All baseline characteristics including demographics, comorbidities, obesity-related factors, and socioeconomic variables were collected at the time a control patient began accruing time in the database at first diagnosis of morbid obesity (BMI >35 kg/m2). Patients were then matched on a 1:1 basis for the propensity to undergo bariatric surgery using propensity scores [17]. Balance was assessed by standardized mean difference (Supplemental Fig. 1) with adequacy considered as less than 20%. McNemar’s test or Wilcoxon signed-rank test was used to compare the matched pairs including for fracture rates. Statistical significance was determined by two-sided α of 0.05. All analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC).

Results

A total of 3439 patients underwent bariatric surgery in the study period; the mean age at the time of surgery was 43 years and 2789 (81.1%) were women. Most (79.4%) of these patients underwent Roux-en-Y gastric bypass (RYGB) while 11.2% had gastric banding and 7.8% had SG procedures. “Other” procedures including BPD and vertical banded gastroplasty comprised only 1.7% of all surgical cases. Fractures occurred in 220 (6.4%) patients (Table 1) at a mean time of 7.6 ± 5.6 years after surgery; 50 (22.7%) of these patients experienced multiple different fracture events while 48 (21.8%) experienced one event resulting in multiple fractured bones. On univariate analysis, patients who experienced a fracture were more likely to have had RYGB or a history of chronic obstructive pulmonary disease, tobacco use, or congestive heart failure. Those who were Caucasian were less likely to experience a postoperative fracture. Most fractures occurred in the upper (39.1%) and lower (33.6%) extremities. Fractures of the hip accounted for 2.3% and facial/skull fractures accounted for 7.3% of all fractures. Fractures at “other” sites including pelvis, spine, ribs, and sternum comprised 17.3% of all recorded fractures.

Table 1 Preoperative characteristics between patients with and without fractures
Full size table

Multivariate logistic regression identified Caucasian race [odds ratio (OR) 0.675; 95% confidence interval (CI) 0.458–0.996], and private insurance (OR 0.698; 95% CI 0.287–0.912) as patient factors that were protective against fractures (Table 2). Tobacco use (OR 3.124; 95% CI 1.633–5.977) and RYGB, when compared to SG (OR 2.175; 95% CI 1.046–4.525), were both independent predictors of fracture.

Table 2 Logistic regression for fracture
Full size table

After 1:1 matching by propensity to undergo surgery, 3880 patients were found to be well matched with propensity score distributions shown in Supplemental Figs. 1 and 2. They were well balanced with all standardized mean differences ≤10% (Supplemental Fig. 3) and similar at baseline except for OSA, which was statistically higher in the surgery group (p = 0.011; Table 3). When comparing the two groups for risk of fracture, the fracture rate was significantly higher in the surgical group than in the non-surgical group (6.4 vs 2.7%, respectively; p < 0.0001). These fractures occurred at an average of 8.2 ± 6.0 years after surgery.

Table 3 Characteristics between matched groups
Full size table

Discussion

In this study of a single institution’s bariatric surgery population of 3439 patients, RYGB and a history of tobacco use were both shown to be independent predictors of fracture. Importantly, we also demonstrated that obese patients undergoing surgery were more than twice as likely to experience a fracture as comorbidity- and BMI-matched patients who did not undergo weight-loss surgery.

Bariatric surgeries that involve a malabsorptive component result in greater weight loss and are more likely to cause protein and various nutritional deficiencies, including vitamin D and calcium deficiency, when compared to purely restrictive procedures [3, 10, 11]. Given the increased risk for these nutritional deficiencies which are known to affect bone health, it follows that patients undergoing malabsorptive procedures are likely at an increased risk for bone loss and fracture. On multivariate analysis of our entire bariatric surgery population, we found that RYGB, a mixed malabsorptive and restrictive procedure, indeed was associated with a higher risk for fracture when compared to SG, a purely restrictive surgery. Guidelines for malabsorptive bariatric surgery recommend postoperative supplementation with vitamin D and calcium and emphasize adequate protein intake in order to minimize these potential deficiencies. They also recommend that all bariatric patients incorporate strength training and daily aerobic activity into their postoperative fitness routines, which can help improve overall bone health [18]. Our analysis also identified tobacco use as an independent risk factor for postoperative fracture after bariatric surgery. Studies have shown that smoking causes decreased bone mineral density and up to one in eight hip fractures in women may be attributable to smoking, according to one meta-analysis [19]. Moreover, smoking is known to impair osteoblast production, causing any sustained fractures to heal at a slower rate [20].

Comparison with other studies of fracture risk

Four studies have been previously published addressing the risk of bone fracture after bariatric surgery. The first was a retrospective cohort study from the UK in which fracture risk was compared between 2079 bariatric surgery patients and 10,442 controls matched for BMI, age, and sex [13]. Lalmohamed et al. found no difference in overall fracture risk between the two groups, though the study only had 2.2 years of follow-up and relatively few observed fractures (n = 39). They also reported no difference in fracture risk between types of bariatric procedures. In their study, 60% of patients underwent gastric banding, a purely restrictive procedure, while only 29% underwent RYGB. It is therefore difficult to compare their results to ours given the difference in practice patterns, with RYGB being far more common in our surgical cohort. Given that restrictive procedures like gastric banding are not associated with alterations in vitamin D and calcium absorption, the majority of patients in the Lalmohamed study were likely at a lower risk for developing skeletal changes and, therefore, fractures.

Another retrospective study of 258 bariatric surgery patients (94% RYGB) in a single county in Minnesota between 1985 and 2004 showed that the fracture risk was two times higher in patients after RYGB compared to age- and sex-matched controls from the general population [14]. The patient population was similar to our cohort in age and gender, though they had a greater proportion of RYGB patients and no SG patients. This is reflective of surgical practice at the time, however, since SG was not yet popular by 2004. Despite this difference, their degree of increase in fracture risk is similar to our analysis and our study identified RYGB as an independent predictor of fracture risk. One weakness of the Nakamura study is that patients were not matched for many confounding factors including comorbid disease or BMI. This may lead to an overestimate of fracture risk in the surgical group, as these patients may have had increased fracture risk as a result of their baseline health and not from undergoing surgery, per se.

A retrospective study from Taiwan comparing fracture risk between 2064 bariatric surgery patients and 5027 propensity score-matched controls showed a 1.2-fold increase in fractures after surgery [15]. Lu et al. reported this increased risk was, as in our findings, limited to malabsorptive procedures; fracture risk was not elevated in patients undergoing purely restrictive bariatric surgery. The matching of patients in this study is similar to ours, except that their patients were not matched based on BMI. Additionally, the Taiwanese cohort was far younger (mean age of 31.8 years) than our bariatric patients, which also may contribute to an underestimated fracture risk.

Finally, Rousseau et al. published a large retrospective, case–control study of 12,676 Canadian bariatric patients who were age and sex matched with 38,028 obese and 126,760 non-obese controls [16]. They showed that bariatric patients had a higher baseline fracture risk in addition to a higher postoperative fracture risk than both the obese [relative risk (RR) 1.38] and non-obese (RR 1.44) control patients, once adjustments were made for demographic data including comorbidities. When comparing fracture risk based on type of bariatric procedure, only patients undergoing BPD had an increased risk of fracture (adjusted RR 1.60). The RYGB group, who, like the BPD group, also underwent a mixed malabsorptive and restrictive procedure, did not have a significantly increased risk of fracture, though there was a trend in this direction. Rousseau et al. therefore reported their findings for RYGB as inconclusive as they had a relatively small number of RYGB cases. Their results may have fallen more closely in line with ours and shown an increased risk of fracture in this population if RYGB had not been underrepresented in their dataset. The Canadian study did not match patients based on BMI, allowing for the possibility that obese controls may have been less obese than the surgical group, which may have led to an exaggeratedly high fracture risk in the surgical cohort. Although obesity has long been thought to be protective against fracture owing to greater bone mineral density, recent studies have called this thought into question [3, 10]. One large meta-analysis showed that women with a higher BMI had an increased risk of fracture (hazard ratio per 1 unit increased in BMI 1.01; 95% CI 1.01–1.02), after adjusting for bone mineral density [21]. We controlled for this possibility by matching for both BMI and comorbidities, and still found the risk of fracture to be elevated after bariatric surgery.

Study strengths and limitations

There are several strengths of the current study including that it is a relatively large group of bariatric patients with a long average follow-up. Additionally, the propensity matching accounted for many variables between groups including BMI, which has only been done in one prior study of fracture risk. This allows a better indication of the fracture risk from surgery itself, and helps eliminate the chance that increased fracture risk is related only to morbid obesity. This study does, however, have several limitations including the risk of bias owing to its retrospective nature. The patient population is comprised only of those who received care at a single institution, and these results are therefore difficult to generalize to a national or international population. We did not have data on patients’ osteopenia or osteoporosis, bone mineral density scans, prior fracture or fall history, baseline activity level or medication use, all of which may affect bone health. Additionally, we did not have information on the etiology of fractures including the level of trauma sustained by patients who experienced a fracture event. Another limitation of this study is that fractures were only recorded within our institution, so patients who were treated for fractures outside of this hospital system may not have been captured.

Conclusions

This study showed that bariatric surgery patients experienced over twice as many fractures as a propensity-matched cohort of morbidly obese patients who did not undergo surgery. Moreover, fracture risk is independently increased in tobacco users and greater in those who undergo mixed malabsorptive and restrictive bariatric procedures when compared to purely restrictive procedures. Bariatric surgery patients should be counseled regarding the increased risk of postoperative fracture, particularly in those who are planning on undergoing procedures with a malabsorptive component. Assistance with tobacco cessation should be provided in the preoperative setting to further decrease the deleterious effects of surgery on bone health. In order to mitigate the effects of bariatric surgery on bone health, physicians must emphasize the importance of patient adherence to guidelines regarding nutritional supplementation and physical activity after surgery.