Low body mass index (BMI) is a recognised risk factor for fracture, particularly at the hip. However, the widely held belief that obesity is protective against fracture has recently been challenged. In a study of postmenopausal women presenting to a Fracture Liaison Clinic with low-trauma fracture, an unexpectedly high prevalence (28%) of obesity was found,1 and in other studies specific associations between obesity and fractures of the ankle,2, 3 upper and lower leg,2 humerus,4 and vertebrae5 have been reported in postmenopausal women. However, with one exception,2, 6 these studies have been relatively small and/or cross-sectional in design.
In this study, we have used a large public health database to examine the association between obesity and incident clinical fracture risk at different skeletal sites in postmenopausal women. We also investigated age-related changes in fracture incidence in obese, overweight, and nonobese women and compared the age at which fractures occur in the three groups.
Materials and Methods
General practitioners (GPs) play an essential role in the public health-care system of Spain, as they are responsible for primary health care, long-term prescriptions, and specialist and hospital referrals. The Spanish public health-care system covers more than 98% of the population. The data in this study were obtained from the Sistema d' Informació per al Desenvolupament de la Investigació en Atenció Primària (SIDIAP) database. The SIDIAP is composed of electronic medical records of a sample of patients attending GPs in Catalonia (northeastern Spain), covering a population of about 5 million patients (80% of the total population for the region) from 274 primary care practices and with a total of 3414 participating GPs.
SIDIAP comprises the clinical and referral events registered by primary care health professionals (GPs and nurses) and administrative staff in electronic medical records, comprehensive demographic information, prescription and corresponding pharmacy invoicing data, specialist referrals, primary care laboratory test results, hospital admissions, and their major outcomes.7 Health professionals gather this information using International Classification of Diseases, 10th revision (ICD-10) codes and structured spreadsheets designed for the collection of variables relevant for primary care clinical management, such as country of origin, sex, age, height, weight, body mass index, smoking and drinking status, blood pressure measurements, preventive care provided, blood and urine test results, etc. Only GPs who achieve quality-control standards can contribute to the SIDIAP database.8 Encoding personal and clinic identifiers ensures the confidentiality of the information in the SIDIAP database.
We identified all female patients aged ≥50 years old in the database in 2009, and who had at least one BMI measurement registered before that year.
Ascertainment of fractures
Fractures registered in 2009 in the SIDIAP database were identified using medical codes for a list of sites of fracture, which are based on the ICD-10 classification (Supplemental Table S1). Fracture sites considered for these analyses were those defined by Center and Eisman9 as major fractures, based on their associated mortality (hip, clinical spine, pelvis, tibia, multiple ribs, and proximal humerus) and the most prevalent minor osteoporotic fracture in our data (wrist/forearm). Data on fracture have been validated in the SIDIAP database using both prospective cohort and hospital admission databases as a reference: Hip, clinical spine, and wrist/forearm fracture coding have been shown to be highly specific (99%, 99%, and 98%, respectively) in SIDIAP.
Body mass index measurements
Only patients with at least one BMI measurement in SIDIAP were included in these analyses. If a patient had more than one BMI measurement registered, we used the value obtained closest to the year 2009, in which data on fractures were collected.
Patients were classified according to the WHO BMI categories10 into three groups: underweight or normal weight (BMI <25 kg/m2), overweight/pre-obese (BMI ≥25 but <30 kg/m2), and obese (BMI ≥30 kg/m2). We combined normal weight and underweight women because the number of women in the underweight group was low (n = 5812), reducing the power to make comparisons with other groups.
Age-specific fracture incidence rates for each fracture site were calculated by dividing the number of patients with a fracture by the total person-years of follow-up and plotted against age (incidence estimates are available from the corresponding author).
Poisson regression models were used to calculate site-specific fracture incidence rates per 100,000 person-years at risk (pyar) and 95% confidence intervals (CIs) for each BMI category and to provide rate ratios (RRs) and corresponding p values to compare fracture rates among overweight and obese versus underweight/normal weight (reference group). Multivariate Poisson models were fitted to adjust for the following potential confounders: age, smoking status, high alcohol intake (at least 4 drinks in a day or ≥7 drinks per week), type 2 diabetes, and oral corticosteroid use for at least 3 months.
As age was normally distributed in patients in the database, mean (standard deviation) age of fracture was estimated for each fracture site and separately for each BMI group, and differences were estimated using independent sample t tests, using the normal/underweight women as the reference group.
All these analyses were two-sided and were carried out using Stata for Windows, version 10.0 (StataCorp, College Station, TX, USA).
In 2009, of the 1,039,878 women aged 50 years or older identified in SIDIAP, 832,775 (80.1%) had a BMI measurement available and were followed up for a median (interquartile range) of 0.995 (0.991 to 0.997) years. They were categorized as: 5,812 (0.7%) underweight, 296,602 (35.6%) normal weight, 266,798 (32.0%) overweight, and 263,563 (31.7%) obese (Fig. 1). The mean (standard deviation) age of participants was 66.9 (12.1) years.
Overall, the most commonly affected sites were wrist/forearm, followed by hip and clinical spine fracture. Age-adjusted hip fracture incidence rates decreased with each higher BMI category: 444/100,000 pyar (95% CI 412 to 479) in the underweight-normal, 315/100,000 pyar (95% CI 294 to 337) in the overweight, and 219/100,000 pyar (95% CI 202 to 238) in the obese. Corresponding RR were 0.71 (95% CI 0.64 to 0.79), p < 0.001 for the overweight group, and 0.49 (95% CI 0.44 to 0.55), p < 0.001 for the obese group. Similarly, clinical spine, wrist/forearm, and pelvis fracture rates were inversely related to BMI, but differences were only significant for the comparison between extreme BMI groups (normal-underweight as reference group compared with obese): age-adjusted RR for clinical spine 0.87 (95% CI 0.77 to 0.98), p = 0.02; RR for wrist/forearm fractures 0.83 (95% CI 0.75 to 0.91), p < 0.001; RR for pelvis fracture 0.43 (95% CI 0.19 to 0.97), p = 0.03. Conversely, the risk of proximal humerus and tibia fractures increased for each higher BMI group, but this was only significant for proximal humerus fracture: age-adjusted RR 1.12 (95% CI 0.98 to 1.29), p = 0.09 for the overweight and 1.23 (95% CI 1.07 to 1.41), p = 0.002 for the obese. Multiple rib fractures were numerically more common in the normal/underweight than in the overweight and obese. However, this association was not linear-shaped (obese women had nonsignificantly higher fracture rates than overweight) and not significant (p = 0.15 for the overweight and p = 0.22 for the obese). Age-specific fracture incidence rates for the studied fracture sites in the BMI groups are shown in Fig. 2.
Further adjustment for other potential confounders (smoking status, high alcohol intake, type 2 diabetes, and oral corticosteroid use) did not change the direction or significance of the observed association between BMI and hip, pelvis, or proximal humerus fractures but attenuated the observed differences in clinical spine and wrist fracture incidence rates across BMI extreme groups (p = 0.30 and p = 0.32, respectively). RR for the age-adjusted and for the multivariate-adjusted effect of BMI on the different fracture sites studied are reported in Table 1.
Table 1. Fracture Incidence Rates for Different Sites and Versus BMI Categories
BMI = body mass index; IR = incidence rate; pyar = person-years at risk; CI = confidence interval; RR = rate ratio.
444.4 (412.1, 479.3)
315.4 (294.6, 337.6)
0.71 (0.64, 0.79); p < 0.001
0.77 (0.68, 0.88); p < 0.001
219.7 (202.1, 238.6)
0.49 (0.44, 0.55); p < 0.001
0.63 (0.55, 0.73); p < 0.001
10.58 (6.48, 17.27)
8.25 (5.48, 12.65)
0.78 (0.39, 1.59); p = 0.45
0.78 (0.63, 0.96); p = 0.017
4.55 (2.39, 8.27)
0.43 (0.19, 0.97); p = 0.03
0.58 (0.47, 0.73); p < 0.001
296.9 (270.7, 325.7)
277.1 (258.6, 296.9)
0.93 (0.83, 1.05); p = 0.25
1.03 (0.87, 1.23); p = 0.70
258.0 (240.1, 277.3)
0.87 (0.77, 0.98); p = 0.02
0.91 (0.76, 1.09); p = 0.30
6.61 (3.56, 12.29)
3.54 (1.96, 6.39)
0.51 (0.18, 1.40); p = 0.15
0.84 (0.38, 3.35); p = 0.40
3.79 (2.24, 6.96)
0.57 (0.21, 1.54); p = 0.22
0.89 (0.36, 1.87); p = 0.49
464.3 (431.2, 499.9)
441.5 (417.7, 466.8)
0.95 (0.87, 1.05); p = 0.29
1.04 (0.92, 1.18); p = 0.53
383.9 (361.8, 407.4)
0.83 (0.75, 0.91); p = 0.0001
0.94 (0.82, 1.06); p = 0.32
203.7 (182.2, 227.8)
229.0 (211.9, 247.3)
1.12 (0.98, 1.29); p = 0.09
1.22 (0.99, 1.49); p = 0.07
250.8 (232.9, 270.0)
1.23 (1.07, 1.41); p = 0.002
1.28 (1.04, 1.58); p = 0.018
30.42 (22.79, 40.62)
31.11 (25.42, 38.80)
1.02 (0.70, 1.50); p = 0.91
1.02 (0.56, 1.86); p = 0.95
34.53 (28.06, 42.62)
1.13 (0.79, 1.66); p = 0.49
1.19 (0.66, 2.15); p = 0.57
Age at fracture was estimated separately for each fracture site and BMI group, and results are shown in Table 2. In our data, obese women with an incident hip fracture were on average 2.7 years (95% CI 1.9 to 3.4; p < 0.001) younger than normal/underweight peers. Interestingly enough, for clinical spine and pelvis fractures, obese women were 1.7 years (95% CI 0.6 to 2.8; p = 0.004) and 3.8 (95% CI 1.1 to 6.5; p = 0.007) years younger, respectively, than normal/underweight women. Conversely, wrist/arm fractures occurred when they were 1.7 years (95% CI 0.8 to 2.6) older than normal/underweight women (p < 0.001). No differences in age at fracture were found for multiple rib, proximal humerus, or tibia fractures.
Table 2. Age at Fracture for Each Skeletal Site and BMI Group
In this population of Spanish women aged ≥50 years, the association between body mass index and fracture was site-dependent: compared with normal and underweight women, overweight women had an almost 25% lower hip fracture risk, and obese women had a 40% lower risk. Pelvis fracture rates were also reduced by about 20% and 40% in the overweight and obese women, respectively. In contrast, the risk of proximal humerus fractures was almost 30% higher in the obese compared with normal/underweight women. In addition, although fracture incidence increased with age in all BMI categories, obese women fractured their hip, spine, and pelvis when they were, on average, 2.7, 1.7, and 3.8 years younger than normal/underweight women, respectively.
Varying associations between fracture site and obesity have been reported in other studies. Consistent with our results, Gnudi and colleagues demonstrated that obesity was associated with a higher risk of proximal humerus fracture and lower risk of hip fracture in a cross-sectional study of 2235 postmenopausal women.4 However, other studies have reported higher rates of ankle,2, 3 lower and upper leg fractures,2 and vertebral fractures2 in obese postmenopausal women. In the current study, tibial fractures showed a nonsignificant trend to be more common in obese women, but ankle fractures were not considered separately. Using data from the Womens' Health Initiative (WHI) study in postmenopausal women, Beck and colleagues reported a significantly greater incidence of lower-extremity fractures in obese versus normal weight women and a significantly lower incidence of hip fracture.11 In men aged ≥65 years, obesity was associated with increased risk of nonvertebral fractures and of hip fracture;12 the latter finding contrasts with the reduced risk of hip fracture in obese postmenopausal women, possibly reflecting different distributions of fat in obese women and men.13, 14
The reasons for site-specific differences in fracture site in obese compared with nonobese individuals have not been established. Obesity is associated with increased risk of falling;11, 15–21 reduced mobility may impair the normal protective responses and, together with the higher body weight, increase the impact of the fall. Furthermore, because of reduced mobility, obese individuals may be more likely to fall backward or sideward rather than forward,22 thus explaining their predilection for upper arm and leg fractures. Increased soft tissue padding may explain the protection against hip and pelvis fractures in women.23
Although bone mineral density (BMD) measurement in obese postmenopausal women with fracture often shows normal or only marginally osteopenic values,1 fractures in the obese population exhibit many of the characteristics of fragility fractures. Thus the frequency of a previous history of fracture is similar in obese and nonobese women with an incident fracture,21 and as shown in the current study, fracture incidence at all sites shows a clear age-related increase. In addition, when compared with their obese counterparts without fracture, obese women with fracture have significantly lower BMD in the spine and proximal femur.2 This may reflect the impaired mobility, poorer self-reported general health, and increased comorbidity reported in obese women with fracture when compared with obese women without fracture.2, 21 In addition, lower cortical volumetric BMD has been reported in women with severe obesity compared with those with lower BMI, possibly as a result of higher serum parathyroid hormone levels.21 Other risk factors for fractures in obese women include early menopause and glucocorticoid use.24
There is no clear explanation why in obese postmenopausal women, hip, spine, and pelvis fractures occurred at a slightly younger age than in nonobese women. Similar findings were also reported in the GLOW study and may reflect the contribution of comorbidities, poor general health, reduced physical mobility, and increased risk of falls to frailty in the obese population.2
Strengths and Limitations
Important strengths of our study are that the SIDIAP database is highly representative for the whole population of our region. The prevalence of obesity in this population (31.7%) is similar to that reported nationally.2 In addition, the sample size of our study population (more than 800,000 women) allowed us to study the association between BMI and different fracture sites separately, in order to demonstrate that this association is not equivalent for all of the studied fractures.
The most important potential limitation of our study is the nonvalidation of fractures at an individual level. It is possible that there are some missing fractures in our data set and that some misclassification may have occurred, particularly with respect to spine and rib fractures. BMI may also have been wrongly reported in some women, although any misclassification is likely to be random. Secondly, fractures of the ankle and upper leg were not specifically explored in these data but were grouped with tibial and hip fractures, respectively. The study was restricted to patients with at least one BMI measurement available and may not be representative of the general population. However, only 20% of the population in the target population had missing information on BMI. Finally, details on bone mineral density, fall frequency, and the degree of trauma associated with fractures were not recorded.
Conclusions and Clinical Implications
The association between obesity and fracture in postmenopausal women is site-dependent: in our cohort, obesity was protective against hip and pelvis fractures but was associated with increased risk of proximal humerus fractures. Fractures in obese women are likely to be associated with a greater morbidity than in nonobese women because of the higher frequency of nonunion,25 greater likelihood of postoperative complications, and slower rehabilitation. Whether currently approved bone-protective therapies reduce fracture risk in obese women is unknown and is an important area for future research.
All the authors state that they have no conflicts of interest.
DPA receives support from the IDIAP Jordi Gol and Institut Catala de la Salut (“4a Convocatòria d'una estada a una Unitat de Recerca de l'IMIM o de l'ASPB”). The Internal Medicine Department and the URFOA IMIM receive support from the RETICEF (Red Temática de Investigación Cooperativa en Envejecimiento y Fragilidad, Instituto Carlos III, Government of Spain). We thank all the health professionals involved in registering data in computerized medical records for SIDIAP.
Authors' roles: Study design: DPA, MP, XN, JC, and ADP. Study conduct: FFA, EH, and DPA. Data management: FFA. Data interpretation: DPA, MP, XN, JC, and ADP. Drafting the manuscript: DPA, MP, JC, and ADP. Revising the manuscript content: All the authors. Approving the final version of the manuscript: All the authors.