Fractures of the femur result mainly from bone fragility. Fractures of the proximal femur are the most frequent, and are considered to be a hallmark of osteoporotic fractures. They are often referred to as “hip fractures” and include the femoral neck and intertrochanteric regions. Subtrochanteric and femoral shaft fractures (ST/FS) are rare occurrences, and are usually related to high-energy trauma in young patients, disease-related fractures, and even more rarely to low-energy falls in elderly people with osteoporosis. Interest in these fractures has increased recently, as many reports have been published suggesting that the distinction between typical and atypical fractures may be related to long-term bisphosphonate use.1–8 However, the increased risk of ST/FS fractures in patients taking bisphosphonates long term has not been clearly established and might be related to fragility fractures.9–12
Although the incidence of hip fractures has been well characterized, the incidence of ST/FS fractures has only recently been reported. A few epidemiology studies have looked at the prevalence of femoral fractures, and pointed out that their occurrence in elderly women might result from bone fragility.13–15 A large-scale, 11-year epidemiological study in the US population aged above 50 years reported an incidence of 3% for the subtrochanteric region and of 5% for the femoral shaft, whereas 87% of fractures were of the proximal femur.16 A higher incidence was found in women than in men. Moreover, the incidence of each type of fracture remained stable over 5 years, but increased exponentially with age. In a retrospective analysis of the Hong Kong hospital database, the incidences of low-trauma subtrochanteric and of femoral diaphysis fractures over a 5-year period17 identified 88 cases of subtrochanteric fractures and 66 of diaphysis fractures, accounting for 3.9% and 2.9%, respectively, of all recorded osteoporotic fractures.17 Although the incidence of ST/FS fractures is lower than that of hip fracture, they account for about 3% of all femoral fractures in the elderly.16, 18 Such fractures occurred at a younger age than hip fractures, and had numerous comorbidity factors.18 There was no difference in the frequency of diseases, such as diabetes, rheumatoid arthritis, chronic pulmonary diseases, heart failure or depression, or the associated prescription drug use between hip and ST/FS.16, 18 The risk factors that lead to these fractures have not been fully identified. Our study set out to compare the incidence of ST/FS fractures to that of hip fractures in an exhaustive French national database, and to identify the risk factors that predispose subjects to ST/FS versus those that predispose for hip fractures.
Materials and Methods
Source of the femoral fracture data: the French National Hospital Discharge Database
Data were obtained from the French National Database, which includes all hospital discharge codes for public and private acute care facilities from 2002 to 2009.19 This database is mandatory and exhaustive, as the funding of each hospital is calculated from this hospital coding. The accuracy and quality of the coding is checked every month in each hospital by a dedicated physician, and an external audit is performed on a yearly basis by physicians of the national medical insurance services. The coding of diagnoses and procedures is based on the international classification of diseases (ICD-10) and the French catalog of procedures (ie, “catalogue des actes médicaux” until 2004 and “classification commune des actes médicaux” from 2004), respectively. The coding rules are updated regularly. Until 2009, the primary diagnosis was defined as the diagnosis that contributed most to the care provided during hospital stays; after 2009, it was related to the reason for admission to the hospital (this change has no consequence in our study, because the primary diagnosis is always the fracture for which the patient is hospitalized for surgical management). Associated diagnoses included all diseases or nonclinical abnormalities that appeared during the patient's stay, followed with complications of chronic diseases, and specific medical conditions that were present before or during the hospitalization for fracture (ie, diabetes, hypertension, kidney transplants, etc.). All procedures performed during the patient's hospital stay were encoded.
From this national database, two extractions were performed for women and men aged 50 years and over: the first was related to hip fractures, and the second to ST/FS fractures (Fig. 1). We extracted all hospital stays related to the management of hip or ST/FS fractures encoded as the primary diagnosis, that is, ICD-10 codes S72.0 and S72.1 for hip, and S72.2 and S72.3 for ST/FS, respectively. Only hospital stays involving a surgical procedure were included. In order to focus on fragility fractures, we excluded any hospitalizations in which cancer or fracture were encoded as a secondary diagnosis (eg, in which a hip fracture was encoded as the primary diagnosis, and ST/FS fracture as a secondary diagnosis, or vice versa), surgical management occurring after the initial surgical treatment (eg, the removal or replacement of a hip prosthesis, removal of osteosynthesis material, reduction of the luxation of a hip prosthesis) in order to exclude fractures that may have occurred at the site of osteosynthesis material or disease-related multitrauma. Here, we could not use, as others, the ICD-10 codes that allowed to know the circumstance of the occurrence of the trauma (codes beginning by “W”).20 This information was underencoded as reported in less than 1% of the selection for 2009, for example. Table 1 described the selected hospitalizations at each step. The final selection consisted of between 80% and 86% and 68% to 80% of all hospitalizations for which hip or ST/FS was encoded as a primary diagnosis, respectively. We counted the hospitalizations for each type of fracture, and stratified by gender and age (50–59, 60–69, 70–79, 80–89, >89 years). We assumed that each hospitalization corresponded to a single patient, and that the likelihood of having a contralateral fracture was very low.
Table 1. Flow Selection of Hospitalizations for Neck/Intertrochanteric and Subtrochanteric/Femoral Shaft
Step 1: Selection of hospitalizations from which primary diagnosis encoded was either femoral neck (S72.0) or intertrochanteric region (S72.1) fractures or subtrochanteric (S72.3)/femoral shaft (S72.4) fracture.
Step 2: Exclusion of hospitalizations with secondary diagnosis mentioned cancer or fracture (eg, in which a hip fracture was encoded as the primary diagnosis, and ST/FS fracture as a secondary diagnosis or vice versa).
Step 3: Exclusion of hospitalization where surgical management occurring after the initial surgical treatment (eg, the removal or replacement of a hip prosthesis, removal of osteosynthesis material, reduction of the luxation of a hip prosthesis).
Step 4: Exclusion of hospitalizations classified in disease-related group of removal of osteosynthesis material or multitrauma.
Overall hospitalization in France
23 781 314
24 300 476
Hospitalizations for femoral neck or intertrochanteric fracture
Hospitalizations for subtrochanteric or femoral shaft fracture
Incidence of femoral fractures according to the fracture site
The incidence of femoral fracture of the hip or ST/FS was classified in women and men, and stratified according to five age classes: 50–59 years, 60–69 years, 70–79 years, 80–89 years, >89 years, for each year from 2002 to 2009.
To calculate the incidence of hip or ST/FS fractures among the general population, we obtained the French civilian population database for each year (from the INSEE Website). We then divided the number of patients with fractures for each given year, age, and gender by the corresponding population for each year adjusted for age and gender. The results were expressed per million inhabitants. We also calculated the adjusted incidence on age using the 2002 population as standard.
Identification of the risk factors associated with ST/FS and hip fractures
The frequency of comorbidity factors was investigated in relation with the incidence of ST/FS and hip fractures respectively in the data obtained in 2009. To this aim, we selected the diseases reported as secondary diagnosis. In France, the definition of the secondary diagnosis corresponded to one of the following situation: any disease or nonclinical abnormalities that appeared during the hospital stay or the follow-up of a chronic disease with or without complication during the hospitalization treated or not or any specific medical conditions (ie, acquired absence of kidney, colostomy status, etc.). Therefore, all the diseases that occur during the hospitalization or that were present before the fractures (diabetes, hypertension, obesity, etc.), are reported. Most of the secondary diagnoses are encoded because the burden of the management is higher and has an impact on financing. We focused on the comorbidity factors that are relevant for fractures, and the more frequent in the general population in addition to age and gender. Comorbidity disorders were reported as secondary diagnoses, and encoded according to ICD-10, that is, dementia (F00.-, F01.-, F02.-, F03.-), dyslipidemia (E78.-), obesity (E66.-), diabetes (E10.-, E11.-, E12.-, E13.-, E.14.-), cardiac arrhythmia (I48, I49.-), and hypertension (I10, I11.-).
To study the difference in the incidence of femoral fractures according to the facture site, a linear trend of gender- and age-specific incidence for each fracture site was identified using the test for trends in proportions.
To assess the frequency of comorbidity factors in hip or ST/FS fractures, obesity, dyslipidemia, diabetes, hypertension, dementia, and cardiac arrhythmias were coded as binary variables (yes/no). Age was dichotomized according to its median value (ie, ≤84 years or >84 years). We first conducted a univariate analysis to study the occurrence of ST/FS factures in relation to gender and age. The same test was used to assess the presence of ST/FS fractures, and the frequency of each of these risk factors. The threshold of a value of p ≤ 0.05 in this analysis was retained for including the variables in the multivariate model. We conducted a multiple logistic regression performed with the general linear model (GLM) function to evaluate the odds ratios (ORs), and 95% confidence intervals (CIs) for the association between ST/FS fractures and risk factors.21 The best fitting and most parsimonious subsets of factors were selected using the stepwise function. The final model included gender, age, dementia, obesity, diabetes, cardiac arrhythmia, and hypertension as explanatory variables. ORs were derived from the coefficients of the logistic regression.
Statistical analyses were performed using the S-PLUS 6.1 statistics package for Windows (Insightful Corporation, Seattle, WA, USA) and STATA 10.1 package for Windows (StataCorp LP, College Station, TX, USA) for the analysis of comorbidities. All p values were two-sided, with a significance level of 0.05.
Data from the French National Hospital Discharge Database
The numbers of hospital stays corresponding to hip and ST/FS fractures in the different age classes are shown for women (Tables 2a and 2b) and for men (Tables 3a and 3b). From 2002 to 2009, the hospitalization rate was higher in women than in men, regardless of the fracture site. The burden of hip fractures was eight times greater than that of ST/FS fractures in both women (from 7.5% to 10.6% depending on the year) and men (from 8.9% to 11.2% depending on the year). From 2002 to 2009, the percentage of ST/FS fractures among fractures of the femur was similar in women and men: 7.5%, 7.5%, 8.2%, 9.2%, 9.8%, 9.9%, and 10.7% in women, and 9%, 8.9%, 9.9%, 10.6%, 10.9%, 10.2%, and 10.3% in men, respectively.
Table 2a. 2002–2009 Evolution of Neck and Trochanteric Fractures in Women
However, from 2002 to 2009, the hospitalization for hip and ST/FS fractures were decreased by 1% and increased by 4.8% and 32%, respectively, in women, and increased by 13% and 32%, respectively, in men. The increase in hospital admissions over this period was greater for ST/FS than for hip fractures.
The incidence in the population depending on the femoral fracture site
The change in incidence is shown for hip (Tables 2a and 3a) and for ST/FS (Tables 2b and 3b) fractures in the different age classes and genders. The incidence of both fracture types increased with age, regardless of the year and gender. In contrast to the change in the number of hospitalizations, the change in the incidence from 2002 to 2009 depended on gender. A decrease from 4368 to 3821 per million was observed in women. In contrast, in men, there was a small variation of the incidence from 1369 to 1479. The changes were statistically significant in both genders, as shown in Tables 2–3. A decrease in the incidence of hip fractures was observed in women in the two highest age classes (80–89 and >89), whereas the incidence in these age classes remained stable for men. In contrast, the incidence of ST/FS fractures per million increased significantly in both genders, from 353 to 412 and from 146 to 168 in women and men, respectively. In both genders, a significant increase was observed in the highest age classes (ie, 70–79, 80–89, and >89 years).
In 2009, the incidence per million of ST/FS fractures in the general population remained low (412 and 168 in women and men aged 50 years and over, respectively). However, women and men aged over 89 years, these levels reached 2966 and 1461, respectively.
Whatever the type of fractures and gender in people aged 50 years and older, we found that the adjusted incidences were slightly lower than nonadjusted incidence and this finding has no impact in the general trends of the 2002–2009 evolution.
Comorbidity factors associated with ST/FS fractures
In the univariate analysis, we found significantly more women (76.1% versus 74.8%, p = 0.023) and more patients with hypertension (34.1% versus 33.0%, p = 0.009) among the patients with hip fracture than those with ST/FS fractures. There were significantly more patients aged 84 years or less (57.7% versus 53.0%, p < 0.0001), with dementia (28.0% versus 24.0%, p < 0.0001), diabetes (10.8% versus 9.9%, p = 0.01), or obesity (3.5% versus 1.6%, p < 0.0001) among the patients with ST/FS fractures than among those with hip fractures. No significant relationship was found between the presence of dyslipidemia in either group of patients (p = 0.39) or between the presence of cardiac arrhythmia (p = 0.15).
We then conducted a multivariate, logistic regression including the variables that had been found to be statistically significant (age, gender, dementia, obesity, diabetes, and hypertension). After performing a stepwise method, the final model included the same variables. Table 4 shows the OR and 95% confidence interval of the risk of having a ST/FS according to a comorbidity factor, after adjustment for the other factors included. According to this model, the risk of having a ST/FS fracture compared with that of having a hip fracture was significantly lower for increased age older than 84 years versus 84 years or younger (p < 0.0001) and for hypertension (p < 0.0001), but was greater for obesity (p < 0.0001) and dementia (p < 0.0001). Gender and diabetes were not associated with the risk of ST/FS fractures.
Table 4. Multiple Logistic Regression Model (Risk of Having ST/FS for Subjects (With the Comorbidity Versus Those Without)
We used the French National Hospital Discharge Database to analyze the incidence of hip and ST/FS fractures from 2002 to 2009. We found that the incidence of hip fractures decreased in women as previously reported19 and a small variation of the incidence per million from 1369 to 1476 in men. A fall in the incidence of hip fractures had already been reported previously for a period of 11 years in women,16, 18 but not consistently in men.16 The reduction in incidence in women might result from improved management of osteoporosis in terms of prevention or medication. However, we also observed an increase in the incidence in men. Such an increase might have been because of a higher frequency of risk factors, changes in bone strength, or changes in the demographic structure of the population, and indeed, previous studies had not analyzed the fracture incidence rates adjusted for demographic changes in the population.22 The adjustment for age and gender in the population for each year make these data more accurate.
In contrast, ST/FS fractures significantly increased in both genders over the course of the same period. Indeed, the incidence increased with age regardless of gender or fracture site, particularly in the older age classes. Moreover, the mean incidence of ST/FS fractures was eight times lower than that of hip fractures, and remained low with the incidence of 412 and 168 per million. Our data differ from previous reports based on hospital databases that showed either a stable incidence of ST/FS fractures in recent years,16 or an increase in women only.18 Methodologic differences may explain these discrepancies, particularly the source and selection of the population, and the expression of the results. In other studies, the incidence was calculated from a representative sample of national American discharges,16, 18 or of American medical insurance data (MarketScan®),16 whereas our data were exhaustive and based on all French hospital discharge reports. The fractures were selected on the basis of ICD-9 codes excluding open fractures, cancer, or Paget disease,16, 18 or encoded as the primary diagnosis and including primary surgical treatments or replacements.18 No significant difference was found between this approach and our selection method. The fact that we used ICD-10 codes rather than ICD-9 codes had no impact on fracture selection. The type of fracture typically included fractures of the hip, subtrochanter and diaphysis.16, 18 In our study, we compared typical hip fractures and ST/FS fractures. In patients aged 50 years and over or 65 years and over, the burden of fractures was expressed using the national discharge rate data and the incidence rate using medical insurance data.16, 18 The last point concerns the choice of the population used, that is, the US 2000 civilian population,16 the midyear census population estimates,18 or the French census for each study year in our study. We found no relevant difference in the evolution of the incidence of ST/FS fracture between adjusted or nonadjusted incidence in women and men aged 50 years and over.
Despite the differences in methodology, we confirmed that the burden of ST/FS fractures is much lower than that of hip fractures. However, in 2009, among the general population we found an incidence per million of ST/FS fractures of 412 in women versus 168 in men. This is higher than in previous reports that were not population-adjusted for the same period: in the Nieves study, which used the US 2000 civilian population,16 the incidence was 0.01 and 0.007% (as it was in the Wang study, which used the midyear census population estimates.18 This might further increase the true financial burden of fragility fractures.23
The prevalence of ST/FS fractures compared with that of hip fractures raises the question of risk factors. On the basis of code collection, of medical history, or of prescriptions, other authors failed to show any difference in the prevalence of chronic diseases or of medications between the two groups.16, 18 Indeed, previous studies were extrapolated from a nationally representative subsample, which could be a source of potential bias. Several chronic diseases or medications were reported that are to be expected in the elderly.24 We found a higher prevalence of dementia and obesity in patients with ST/FS than in those with hip fractures. Dementia is a risk factor for hip fractures,25 as sufferers are prone to falls and to osteoporosis.26, 27 A similar frequency of dementia was reported in hip and ST/FS fractures.16 In our study, the higher frequency of dementia could be related to the exhaustive nature of the reports. Obesity is often found to be a protective factor against hip fracture;28, 29 however, excess subcutaneous fat with less lean mass may alter the gait, thus predisposing individuals to falls. Indeed, obesity modifies the mechanical impact at the hip, which may consequently be transmitted to the diaphysis,30 where the adipose tissue may have less shock-absorbing effect than at the hip. We also found that higher age is protective against ST/FS fractures. The higher incidence reported at younger ages might indicate that the traumatic nature of such fractures makes them more likely to occur in a population with a greater propensity to trauma.
Despite a slightly increased risk, cases of ST/FS fracture are rare and the absolute risk remains low as described previously.11 Despite the exhaustive nature of our data collection, our study has some limitations. The French Hospital National Discharge Data did not include the medical consumption of treatments related to osteoporosis, and no data was available for outpatients. The characteristics of the fracture trait is not encoded in the ICD-10 codes (or in the ICD-9 codes), which makes it impossible to distinguish between typical and atypical fractures. Moreover, the condition of the fracture is not provided, and therefore traumatic fractures could have been included, whereas traumatic DRG was excluded. Another point concerned the comorbidities assessed from 2009. This corresponded to medical chronic diseases and to the definition of secondary diagnosis. Despite the fact that our data are anonymous, which does not allow verifying in the medical records, we can expect the exhaustivity of their encoding because the hospitalization of patient with comorbidies has a high impact in terms of management and financing. Only the diagnoses before or during the hospitalization are reported, and not those that might occur after as our system is not linked to outpatient files as it is in other countries such as in North of Europe or North America.
In conclusion, we found that the incidence of hip fractures decreased in women and increased in men over a period of 8 years. The incidence of ST/FS fractures increased significantly in both genders throughout the same period, but remained low at the population level. The cause of such as increase remains to be determined in the light of the risk factors.
All authors state that they have no conflicts of interest.
Authors' roles: Study design: MM and MCS. Study conduct: MM, MCS, and AO. Data analysis and interpretation: MM, MCS, and AO. Drafting and revising the manuscript: MM, MCS, and AO. MM takes the responsibility for the integrity of the data analysis.