SEARCH

SEARCH BY CITATION

Keywords:

  • BISPHOSPHONATE;
  • SUBTROCHANTERIC AND FEMORAL SHAFT FRACTURES;
  • ATYPICAL FRACTURE

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References
  10. Supporting Information

The case definition, community incidence, and characteristics of atypical femoral shaft fractures (FSFs) are poorly understood. This retrospective study utilized electronic medical records and radiograph review among women ≥50 years of age and men ≥65 years of age from January 1996 to June 2009 at Kaiser Permanente Northwest to describe the incidence rates and characteristics of subgroups of femur fractures. Fractures were categorized based on the American Society for Bone and Mineral Research (ASBMR) as atypical fracture major features (AFMs) (low force, shaft location, transverse or short oblique, noncomminuted) and AFMs with additional minor radiograph features (AFMms) (beaking, cortical thickening, or stress fracture). There were 5034 fractures in the study. The incidence rates of FSFs (without atypical features) and AFMs appeared flat (cumulative incidence: 18.2 per 100,000 person-years, 95% CI = 16.0–20.7; 5.9 per 100,000 person-years, 95% CI = 4.6–7.4; respectively) with 1,271,575 person-years observed. The proportion of AFMs that were AFMms increased over time. Thirty percent of AFMs had any dispensing of a bisphosphonate prior to the fracture, compared to 15.8% of the non-atypical FSFs. Years of oral glucocorticosteroid dispensing appeared highest in AFM and AFMm fractures. Those with AFMs only were older and had a lower frequency of bisphosphonate dispensing compared to those with AFMms. We conclude that rates of FSFs, with and without atypia, were low and stable over 13.5 years. Patients with only AFMs appear to be different from those with AFMms; it may be that only the latter group is atypical. There appear to be multiple associated risk factors for AFMm fractures. © 2012 American Society for Bone and Mineral Research.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References
  10. Supporting Information

Osteoporosis prevalence is increasing with an aging U.S. population1, 2 and resulted in approximately 2 million new fractures in the U.S. in 2005, costing $17 billion. Costs are projected to increase ∼50% by 2025.2 Bisphosphonates, the most commonly prescribed treatment for osteoporosis, reduce the risk of vertebral and nonvertebral fractures in patients with preexisting vertebral fractures and low bone mineral density (BMD),3 with BMD-confirmed osteoporosis,4–6 and in those with a prior hip fracture.6 Recently,3–5, 7 however, case reports have emerged of an unusual type of femoral fracture in patients using oral bisphosphonates.8–12

Nevasier and colleagues8 described these as low-energy fractures of the femoral shaft with a simple transverse pattern and hypertrophy of the diaphyseal cortex. Many of these fractures also included a medial cortical spike “unicortical beak.” A retrospective review of this convenience sample of low-energy femoral shaft fractures (FSFs) (n = 70)8 found that 25 (36%) were taking bisphosphonates (all alendronate). Nineteen (76%) of those taking bisphosphonates had the characteristic fracture, compared to one patient (2%) not taking bisphosphonates. Other smaller case series reported subtrochanteric fractures with lateral cortical thickening suggestive of stress fracture, with a high proportion reported in alendronate users.9, 10 Some have suggested that these fractures may be due to suppression of bone turnover, leading to accumulation of microdamage in some long-term users of oral bisphosphonates.11

However, it is not possible using case reports alone to compare the potential risks of bisphosphonate use to the risk of these fractures, because case reports may be biased as a result of limited historical patient information, the use of varying definitions of atypical FSFs, and having unclear or no comparison groups. A recent analysis of three randomized bisphosphonate trials3 did not show a significant relationship between bisphosphonates and FSFs, although incidence rates and numbers of fractures were low and confidence intervals (CIs) were wide. Furthermore, this study could not identify atypical features from radiographs. There are limited databases available for describing population-based fracture rates and radiographic features of atypia.

Therefore, we studied incidence rates and fracture characteristics of femoral fractures, with and without atypia, over a 13.5-year period using electronic medical records (including radiographs) from Kaiser Permanente Northwest (KPNW), a health maintenance organization (HMO) with membership representative of the local region.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References
  10. Supporting Information

Study design overview and population

Part 1 was a retrospective cohort study that utilized electronic medical-record (EMR) data and dual orthopedic expert radiograph review to determine incidence rates of subcategories of femur fractures among women ≥50 years of age and men ≥65 years of age who were KPNW members at any time from January 1, 1996 through June 30, 2009. Part 2 was a cross-sectional descriptive analysis that characterized fractures reviewed radiographically.

Study site and data sources

The study was conducted at KPNW, an HMO that has ∼480,000 members similar in age, gender, ethnicity, and economic status to the local population.3 The HMO has used an integrated EMR since 1996 and stores radiographs and documents containing radiographic interpretations. BMD measurements (when available), generated using dual X-ray absorptiometry (DXA),13 include total BMD of the hip and lumbar spine (L1–L4). The study design and procedures were approved by the study site's Institutional Review Board, which waived the need for informed consent.

Study variables and outcome review

The primary outcome was femoral fracture. All qualifying fractures were included in cases where patients had multiple fractures. We used electronic data and an algorithm to identify possible new femoral fractures by extracting any evidence of a hospitalization or follow-up care for International Statistical Classification of Diseases and Related Health Problems, 9th edition (ICD-9) fracture codes (Supplemental Appendix 1). Possible fractures were first subcategorized for review, based upon ICD-9 codes, into femoral neck, intertrochanteric or unspecified trochanteric, subtrochanteric or shaft, or distal femur. Expert medical record reviewers determined if the fracture was new and categorized the fracture site using notes from an orthopedic expert (operative report, hospital admission, visit, and discharge reports) and radiology reports. We excluded cases of old fractures, cases in which there was a tumor found at the fracture site, cases of Paget's disease of bone, or cases in which the fracture occurred at an implant or prosthesis site. We assigned the day of the fracture (or date of first treatment if fracture date was unknown) as the event date.

All qualifying FSFs, distal femur fractures, and a random sample of 300 femoral neck and 300 intertrochanteric fractures were sent to a trained orthopedic physician blinded to patient treatment data for review. Additional femoral neck (n = 53) and/or intertrochanteric fractures (n = 27) were included for radiographic review if the patient also had an FSF. All FSFs and distal femur fractures were dual-reviewed by two orthopedic experts, blinded to each other's results. For cases in which there was disagreement about a potential atypical fracture case definition,14 the radiographs were reviewed again, this time by both orthopedic reviewers together. If no consensus could be reached, the radiograph was evaluated by a third orthopedic physician reviewer. Supplemental Appendix 2 details missing radiographs by year.

FSFs were further categorized based on criteria defined by the American Society for Bone and Mineral Research (ASBMR) into those with atypical fracture major features (FSF/AFM; femur fracture from slightly distal to lesser trochanter to slightly proximal to the supracondylar flare, associated with no or minimal trauma, transverse or short oblique morphology, and noncomminuted) and those without these features (non-atypical FSF). FSF/AFMs were further categorized into those with additional minor radiograph features (FSF/AFMm; localized periosteal reaction of the lateral cortex [beaking], increase in cortical thickness of the diaphysis, or signs of stress [unicortical] fracture). Orthopedic interrater reliability was assessed after the first 50 dual-reviewed FSFs. The kappas for features of an atypical femur fracture ranged from 0.62 for any atypical feature to 0.84 for fractures with transverse morphology.

Because not every fracture had a detailed medical record to review or a radiograph to interpret, we assigned final fracture location for the incidence analyses based upon the highest quality data available; ie, in order of radiograph, chart review, then ICD-9 code.

Other study variables

Additional covariates included age, gender, length of membership, other comorbidities, and medications of interest. Osteoporosis risk-related measures or other conditions of interest included the BMD T-score (during the 36 months prior to event date), prior osteoporotic fracture, hypothyroidism, diabetes mellitus, chronic kidney disease, chronic obstructive pulmonary disease/asthma, depression (in the 12 months prior to event), mean weight, weight < 127 pounds (57.6 kg), and body mass index (BMI) using the most recent weight in the 24 months prior to the event and any available height, and current smoking status. Conditions or medications associated with risk of falls15 included osteoarthritis, rheumatologic conditions, dementia in the 12 months prior to index, and any dispensing of benzodiazepines or tertiary tricyclic antidepressants15 or other medications of interest (proton pump inhibitors and levothyroxine) during the 12 months prior to fracture. As a measure of disease burden,16 we determined the mean number of unique generic drugs dispensed for each participant during the 12 months prior to index.

We assessed oral glucocorticosteroid and osteoporosis medication (bisphosphonates, estrogen, calcitonin, raloxifene) dispensing (Tables 2 and 3), and categorized fractures into no or minimal trauma (fall from standing height or less) versus higher force (falls from more than a standing height, or non-fall causes such as crush injury and severe trauma). An orthopedic reviewer assessed the presence or absence of prodromal pain in atypical fractures using EMR data after radiograph reviews were completed.

Analysis

Continuous data were summarized as mean ± SD. We determined incidence rates per year and CIs of femoral fracture subgroups including atypical fractures. Ninety- five percent CIs were constructed based on a Poisson distribution. We calculated length of membership for the entire study population, to determine population-level person-years of observation (denominator); person-time from January 1, 1996 through June 30, 2009 for each active member from the time they were at least 50 years of age (women) or 65 years of age (men), until they: (1) left KPNW; (2) died; or (3) the observation period ended. The numerator of fracture incidence was the number of each type of fracture, either by year or over the entire study (cumulative incidence). Among those with at least 12 months of continuous HMO membership prior to the fracture, we also assessed frequency or mean values of the other variables of interest. Odds ratios (ORs) and 95% CIs were calculated to compare the odds of bisphosphonate, glucocorticosteroid, and proton pump inhibitor dispenses between fracture categories. For the ORs, we calculated unadjusted and adjusted (for age, gender, glucocorticosteroid dispensing, and number of medications) ORs.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References
  10. Supporting Information

Figure 1 shows the flow used to generate the study fractures. After exclusions, there were 5034 new fractures included in the incidence rate analyses and 864 cases that had radiographs reviewed for fracture subgroup analyses. Supplemental Appendix 3 details fracture recategorization at each review level. Radiographic and other atypical fracture case definition features of FSFs are described in detail in Table 1.

thumbnail image

Figure 1. Study population flow.

Download figure to PowerPoint

Table 1. Radiographic and Other Case Definition Features of Atypical FSF
FSF subcategoryFSF totalFSF (not atypical)FSF (atypical)
  • ASBMR atypical fracture major features: femur fracture from just distal to lesser trochanter to just proximal to the supracondylar flare, associated with no or minimal trauma, transverse or short oblique morphology, noncomminuted; may have minor radiograph features of localized periosteal reaction of the lateral cortex (beaking), increase in cortical thickness of the diaphysis or signs of stress fracture.

  • ASBMR = American Society for Bone and Mineral Research; FSF = femoral shaft fractures.

  • a

    Categories are not mutually exclusive.

  • b

    Category includes: butterfly fragment, isolated lateral condyle injury, no discrete fracture line, bone edema of unknown origin.

n (%)197 (100)122 (62)75 (38)
Major radiographic features
 Morphology, n (%)a
  Transverse (<30°)45 (22.8)10 (8.2)35 (46.7)
  Short oblique (30° to 60°)50 (25.4)7 (5.7)43 (57.3)
  Long oblique (>60°)101 (51.3)98 (80.3)3 (4.0)
  Otherb6 (3.0)6 (4.9)0 (0.0)
  Comminuted16 (8.1)16 (13.1)0 (0.0)
Minor radiographic features, n (%)
 Any cortical thickening32 (16.2)13 (10.7)19 (25.3)
 Generalized11 (5.6)1 (0.8)10 (13.3)
 Stress (unicortical) fracture7 (3.6)2 (1.6)5 (6.7)
 Beaking
  Possible/definitive26 (13.2)9 (7.4)17(22.6)
No or minimal trauma, n (%)175 (88.8)100 (82.0)75 (100)

Of the 197 FSFs with radiographs, 122 were not atypical and 75 met at least the major criteria for atypia. Typical findings could only be assessed among the 75% of FSFs that had radiographs.

We observed a total of 1,271,575 person-years during the study period, which included an average of 98,580 people per year. Figure 2 shows the incidence rates of fracture subtypes. Overall, FSF rates appear to be flat. The incidence rate of atypical FSFs with major features (FSF/AFMs) has a flat appearance and the mean incidence for the study period was 5.9 per 100,000 person-years (95% CI, 4.6–7.4). The incidence rate of atypical FSFs with major and additional minor radiograph features (FSF/AFMms) is “close to zero” until 2000, then has a low, slightly increasing appearance. Hip fractures appeared to decline and incidence rates of non-atypical FSF and distal femur fractures remained flat. Fracture characteristics across the major fracture subcategories with radiographic reviews are described in Table 2.

thumbnail image

Figure 2. (A) Incidence rates of femoral neck intratrochanteric and distal 1/3 femur or above the knee and overall subtrochanteric or femoral shaft fractures. Incidence rates are based on best available evidence for fracture site. *Distal femur means distal to condylar flare. †Rates are graphed on a different scale than the hip fractures. (B) Incidence rates of subtrochanteric or femoral neck fracture subtypes based on radiograph review. *Major features of ASMBR atypical fractures: femur fracture from just distal to lesser trochanter to just proximal to the supracondylar flare, associated with no or minimal trauma, transverse or short oblique morphology, noncomminuted. **Adds minor radiograph features of localized periosteal reaction of the lateral cortex (beaking), increase in cortical thickness of the diaphysis, or signs of stress fracture.

Download figure to PowerPoint

Table 2. Characteristics of Fractures With Radiograph Review and ≥12 Months Health Plan Membership
 OverallFemoral neckIntertrochantericDistal femurFemoral shaftFemoral shaft (not atypical)aFemoral shaft (atypical)a
  • Number (N) in square brackets represents the number of fractures for which values of the parameter are available.

  • ASBMR = American Society for Bone and Mineral Research; BMI = body mass index; COPD = chronic obstructive pulmonary disease.

  • a

    ASBMR atypical fractures major features: femur fracture from just distal to lesser trochanter to just proximal to the supracondylar flare, associated with no or minimal trauma, transverse or short oblique morphology, noncomminuted; may have minor radiograph features of localized periosteal reaction of the lateral cortex (beaking), increase in cortical thickness of the diaphysis or signs of stress fracture.

  • b

    Among those with ≥1 dispenses.

  • c

    Omeprazole (Prilosec), lansoprazole (Prevacid), rabeprazole (Aciphex), pantoprazole (Protonix), esomeprazole (Nexium), and Zegarid (rapid release form of omeprazole).

All fractures, n86420732913119712275
Fractures with ≥12 months health plan membership, n (%)803 (92.9)193 (93.2)302 (91.8)124 (94.7)184 (93.4)114 (93.4)70 (93.3)
 Age, years, mean (SD)80.2 ± 10.381.8 ± 8.682.7 ± 9.475.5 ± 11.677.7 ± 11.078.1 ± 10.477.0 ± 11.9
 Female, n (%)629 (78.3)146 (75.7)237 (78.5)113 (91.1)133 (72.3)77 (67.5)56 (80.0)
 Membership length, years, mean (SD)21.5 ± 13.821.5 ± 14.722.9 ± 14.219.6 ± 12.620.4 ± 13.021.4 ± 13.218.8 ± 12.4
 Race, white, n (%)[N = 781] 750 (96.0)182 (96.3)289 (97.0)114 (99.1)165 (92.2)[N = 112] 104 (92.9)[N = 67] 61 (91.0)
 BMD value available in the 3 years prior to the fracture, n (%)63 (7.8)13 (6.7)19 (6.3)12 (9.7)19 (10.3)9 (7.9)10 (14.3)
 T-score hip, mean (SD)[N = 52] −2.0 ± 1.2[N = 11] −1.5 ± 0.7[N = 16] −2.3 ± 0.9[N = 9] −2.4 ± 1.1[N = 16] −1.9 ± 1.5[N = 9] −2.2 ± 0.9[N = 7] −1.7 ± 2.1
 T-score spine, mean (SD)[N = 63] −1.6 ± 1.7[N = 13] −0.9 ± 1.2[N = 19] −2.1 ± 1.4[N = 12] −1.6 ± 1.8[N = 19] −1.6 ± 2.4[N = 9] −1.2 ± 1.5[N = 10] −1.9 ± 2.5
 History of fracture, n (%)101 (12.6)14 (7.3)51 (16.9)15 (12.1)21 (11.4)12 (10.7)9 (12.9)
 Weight, lbs, mean (SD)[N = 740] 148.5 ± 41.1[N = 180] 138.9 ± 29.0[N = 286] 139.4 ± 35.2[N = 106] 166.6 ± 56.4[N = 168] 163.0 ± 42.3[N = 103] 165.7 ± 42.5[N = 65] 158.8 ± 41.8
 Weight <127 lbs (57.6 kg), n (%)246 (33.2)67 (36.7)118 (41.3)28 (26.4)34 (20.2)20 (19.4)14 (21.5)
 BMI, kg/m2, mean (SD)[N = 626] 26.0 ± 6.7[N = 153] 23.8 ± 4.3[N = 240] 24.5 ± 5.3[N = 87] 30.6 ± 9.9[N = 146] 28.0 ± 6.8[N = 91] 28.0 ± 7.2[N = 55] 28.0 ± 6.1
 Smoking, n (%)61 (7.6)10 (5.2)29 (9.6)8 (6.5)14 (7.6)10 (8.8)4 (5.7)
 Conditions associated with osteoporosis or falls, n (%)
  COPD/asthma156 (19.4)36 (18.6)75 (24.8)19 (15.3)26 (14.1)14 (12.3)12 (17.1)
  Chronic kidney disease91 (11.3)20 (10.4)34 (11.3)17 (13.7)20 (10.9)13 (11.4)7 (10.0)
  Dementia123 (15.3)36 (18.7)59 (19.5)10 (8.1)18 (9.8)10 (8.8)8 (11.4)
  Depression138 (17.2)32 (16.6)51 (16.9)25 (20.2)30 (16.3)18 (15.8)12 (17.1)
  Diabetes mellitus171 (21.3)33 (17.1)52 (17.2)37 (29.8)49 (26.6)38 (33.3)11 (15.7)
  Hypothyroidism89 (11.1)24 (12.4)36 (11.9)11 (8.9)18 (9.8)13 (11.4)5 (5.7)
  Osteoarthritis189 (23.5)41 (21.2)86 (28.5)24 (19.4)38 (20.7)19 (16.7)19 (27.1)
  Rheumatologic conditions29 (3.6)7 (3.6)7 (2.3)2 (1.6)13 (7.1)8 (7.0)5 (5.7)
 Disease/medication burden
  Number of medications, mean (SD)9.9 ± 7.110.0 ± 6.79.8 ± 6.99.9 ± 7.310.0 ± 7.610.2 ± 7.89.6 ± 7.3
 Medications associated with fracture risk
  Benzodiazepines, n (%)36 (4.5)10 (5.2)9 (3.0)10 (8.1)7 (3.8)3 (2.6)4 (5.7)
  Antidepressants (tertiary tricyclic), n (%)53 (6.6)13 (6.7)21 (7.0)8 (6.5)11 (6.0)7 (6.1)4 (5.7)
  Glucocorticosteroids, n (%)
   Any dispensings past 12 months, n (%)113 (14.1)28 (14.5)46 (15.2)11 (8.9)28 (15.2)19 (16.7)9 (12.9)
   Years of ≥1 dispensings, mean (SD)b[N = 300] 2.7 ± 2.5[N = 71] 2.6 ± 2.5[N = 117] 2.7 ± 2.4[N = 44] 2.0 ± 1.5[N = 68] 2.9 ± 3.0[N = 43] 2.7 ± 2.9[N = 25] 3.3 ± 3.1
    Years of days dispensed ≥90 days, mean (SD)b[N = 48] 3.0 ± 2.4[N = 9] 2.8 ± 2.2[N = 21] 2.6 ± 2.0[N = 5] 1.8 ± 1.3[N = 13] 4.2 ± 3.1[N = 9] 3.8 ± 3.4[N = 5] 5.3 ± 2.5
  Anticonvulsants, n (%)78 (9.7)19 (9.8)30 (9.9)10 (8.1)19 (10.3)13 (11.4)6 (8.6)
  Proton pump inhibitorsc157 (19.6)32 (16.6)55 (18.2)26 (21.0)44 (23.9)27 (23.7)17 (24.3)
  Levothyroxine138 (17.2)32 (16.6)52 (17.2)23 (18.6)31 (16.8)21 (18.4)10 (14.3)
 Osteoporosis treatments
  Estrogen
   Any dispensings past 12 months, n (%)73 (9.1)15 (7.8)30 (9.9)11 (8.9)17 (9.2)11 (9.7)6 (8.6)
     Years of ≥1 dispensings, mean (SD)b[N = 255] 4.7 ± 3.0[N = 60] 5.4 ± 3.1[N = 98] 4.7 ± 2.9[N = 39] 3.9 ± 2.6[N = 58] 4.4 ± 3.0[N = 3] 4.2 ± 2.9[N = 21] 4.9 ± 3.2
  Bisphosphonate
   Any dispensing past 12 months, n (%):129 (16.1)29 (15.0)56 (18.5)17 (13.7)27 (14.7)10 (8.8)17 (24.3)
   Any dispensings at any time prior to event, n (%):178 (22.2)38 (19.7)76 (25.2)25 (20.2)39 (21.2)18 (15.8)21 (30.0)
    Years of ≥1 dispensings, mean (SD)[N = 178] 3.1 ± 2.3[N = 38] 2.8 ± 2.2[N = 76] 2.8 ± 1.8[N = 25] 2.8 ± 2.0[N = 39] 3.9 ± 3.1[N = 18] 3.4 ± 2.6[N = 21] 4.4 ± 3.5
    ≥5 Years of use among all fractures, n (%)b[N = 803] 35 (4.4)[N = 193] 6 (3.1)[N = 302] 11 (3.6)[N = 124] 5 (4.0)[N = 184] 13 (7.1)[N = 114] 6 (5.3)[N = 70] 7 (10.0)

Over 90% (92.9%) of the study population had ≥12 months of membership. Only 13.8% of the all fractures and 12.0% of FSF had <5 years of membership (data not shown). Among those with ≥12 months of membership, the mean age in years was 80.2 ± 10.3; mean age of those with hip fractures was higher (81.8 ± 8.6years) than those with FSFs (77.7 ± 11.0 years). The group was 78.3% female and 96.0% white, and findings were similar across subgroups. Only 7.8% had a BMD value available in the 3 years prior to the fracture.

The mean number of medications dispensed in the 12 months prior to fracture was 9.9 ± 7.1; a small percentage of patients had received benzodiazepines (4.5%) or tertiary tricyclic antidepressants (6.6%), which are both associated with increased risk of falls. An oral glucocorticosteroid was dispensed in 14.1% of patients in the 12 months prior to fracture, with the atypical fracture group having the highest absolute mean observed number of years of any dispensing (3.3 ± 3.1 years) and number of years with ≥90 days supplied (5.3 ± 2.5 years). Nearly 10% (9.7%) were dispensed an anticonvulsant medication in the year prior to fracture.

Dispensings of calcitonin and estrogen were similar across groups and raloxifene dispensing was rare (data not shown). Overall, 16.1% had a dispensing of a bisphosphonate in the 12 months prior to fracture; there was a mean of 3.1 ± 2.3 years of at least one dispensing and 19.7% of the population had at least one dispensing in ≥5 study years. Of the atypical FSF group, 24.3% had a bisphosphonate dispensing in the year prior to fracture. Of those, the mean number of years for those with ≥1 years of dispensing was 4.4 ± 3.5 years and 33% had ≥5 years of use. There were similar frequencies of dispensings of proton pump inhibitors and levothyroxine across groups.

The fracture characteristics across the subcategories of atypical subtrochanteric fractures and FSFs (AFM only [Supplemental Fig. 1A] versus AFMm radiograph features: prefracture [Supplemental Fig. 1B] and postfracture [Supplemental Fig. 1C]) are shown in Table 3.

Table 3. Comparison of Characteristics of Non-Atypical FSFs to Atypical FSFs With Major Criteria Only (AFM) and Atypical FSFs With Major and Minor Criteria (AFMm)
 Subtrochanteric or FSF (non-atypical FSF)Atypical FSF with major criteria only (AFM)Atypical FSF with major and minor criteria (AFMm)
  1. Number (N) in square brackets represents the number of fractures for which values of the parameter are available. ASBMR atypical FSF major criteria: femur fracture from just distal to lesser trochanter to just proximal to the supracondylar flare, associated with no or minimal trauma, transverse or short oblique morphology, noncomminuted. ASBMR atypical FSF minor criteria: localized periosteal reaction of the lateral cortex [beaking], increase in cortical thickness of the diaphysis, signs of stress fracture.

  2. AFM = atypical FSF with major criteria; AFMm = atypical FSF with major and minor criteria; ASBMR = American Society for Bone and Mineral Research; BMD = bone mineral density; BMI = body mass index; FSF = femoral shaft fracture; N/R = not reviewed.

All fractures, n1225322
Fractures with ≥12 months membership, n (%)114 (93.4)49 (92.5)21 (95.5)
 Age, years, mean (SD)78.1 ± 10.479.8 ± 10.870.5 ± 12.0
 Female, n (%)77 (67.5)37 (75.5)19 (90.5)
 Membership length, years, mean (SD)21.4 ± 13.219.7 ± 12.716.8 ± 111.8
 Race, white, n (%)[N = 112] 104 (92.9)[N = 47] 43 (91.5)[N = 20] 18 (90.0)
 BMD value available in the 3 years prior to the fracture, n (%)9 (7.9)2 (4.1)7 (33.3)
 T-score total hip, mean (SD)[N = 9] −2.2 ± 0.9[N = 2] −4.2 ± 0.4[N = 5] −0.7 ± 1.6
 T-score spine L1–L4, mean (SD)[N = 9] −1.2 ± 1.5[N = 2] −4.0 ± 0.7[N = 7] −1.0 ± 2.4
 Weight, lbs, mean (SD)[N = 103] 165.7 ± 42.5[N = 46] 159.6 ± 43.1[N = 18] 156.9 ± 40.6
 BMI, kg/m2, mean (SD)[N = 91] 28.0 ± 7.2[N = 37] 27.9 ± 6.4[N = 20] 28.0 ± 5.4
 Disease/medication burden, number of medications, mean (SD)10.2 ± 7.89.5 ± 7.59.8 ± 7.3
 Oral glucocorticosteroids, years of use, mean (SD)[N = 43] 2.7 ± 2.9[N = 17] 2.6 ± 2.6[N = 8] 4.8 ± 4.1
 Osteoporosis treatments, bisphosphonate, any dispenses past 12 months, n (%)10 (8.8)6 (12.2)11 (52.4)
  Any dispense any time prior to index date, n (%)18 (15.8)8 (16.3)13 (61.9)
   Years of use, mean (SD)[N = 18] 3.4 ± 2.6[N = 8] 2.5 ± 1.6[N = 13] 5.6 ± 3.9
   ≥5 Years of use among all fractures, n (%)[N = 114] 6 (5.3)[N = 49] 1 (2.0)[N = 21] 6 (28.6)
 Prodromal painN/R0 (0.0)6 (27.3)

Most characteristics were similar across these two atypical fracture definitions, except that those with AFM only appeared older, to perhaps have lower BMD, and to have fewer bisphosphonate dispensings than those with fractures with AFMm radiograph features.

We also computed the unadjusted and adjusted ORs (95% CI) of ever having a bisphosphonate dispensed in atypical FSFs versus non-atypical FSFs (unadjusted OR 2.29 [95% CI, 1.12–4.68]; adjusted OR 2.11 [95% CI, 0.99–4.49]) The ORs for exposure to glucocorticosteroids and proton pump inhibitors were also explored between these fracture categories and were not significantly different between the groups (data not shown).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References
  10. Supporting Information

We observed a low and largely stable rate of atypical FSF that met the ASBMR major criteria across the study period, with a mean of 5.9 per 100,000 person-years (95% CI, 4.6–7.4). Most previously published reports on FSF could not determine an atypical FSF incidence rate because they lacked radiographs and could not longitudinally observe a representative community of bisphosphonate users and nonusers.3, 17, 18 This study confirmed femoral fractures subgroups through chart and radiograph review spanning a 13.5-year period (including years when bisphosphonates were rarely used) and was able to determine the observed number of person-years at risk in a community. This is important because using administrative data may overestimate fractures due to codes that appear for old fractures; also, fracture location could be miscoded. Of the 197 subtrochanteric FSFs classified by our orthopedic review of radiographs, only 130 (66%) had been correctly identified by ICD-9 codes.

The FSF incidence rates found here are nearly identical to those found by Black and colleagues,3 who performed secondary analyses on three randomized bisphosphonate trials among women >65 years of age and found an FSF rate of 23 per 100,000 patient-years by using radiograph reports but not the actual radiographs. That study evaluated only women with osteoporosis so rates may have been inflated. Although Giusti and colleagues12 used a slightly different definition of atypical fractures, their findings were similar. In 2008, Schilcher and colleagues19 found an atypical fracture rate of 0.9 per 100,000 person-years in Sweden among those who had not used bisphosphonates, compared to 55 per 100,000 person-years among bisphosphonate users, with a relative risk of 47.3 among ever-users. One of the limitations of that study was that it was conducted during 1 year only and thus researchers were unable to assess trends in atypical fractures. In a case-control portion of the study, 78% of atypical FSF and 10% of the non-atypical FSF had received bisphosphonates in the prior 3 years (OR, 33.3; 95% CI, 14.3–77.8).19 This OR is much higher than that found in the present study. This difference may be explained by our difference in methods (our study is not a true case-control study), varying interpretations of what constitutes an atypical fracture, and by the larger number of FSFs and longer period of drug dispensing history that we were able to observe. Giusti and colleagues20 compared current bisphosphonate use in 10 atypical FSFs compared to 63 non-atypical FSFs (OR, 17.0; 95% CI, 2.6–113.3). Although this OR was also larger than what we observed, the study considered only cases meeting minor criteria to be atypical, the numbers were small and CIs were large. Shorter periods of medication dispensing records observed could have led to misclassification of patient medication exposure.

We were able to compare characteristics of femoral fractures across fracture subgroups. This technique allows for comparisons of subgroups to be made within a patient population at similar risk for osteoporosis, which may mitigate possibilities for introduction of confounding. The highest number of bisphosphonate dispensings and mean number for years of use of glucocorticosteroids was in the atypical fracture subgroup, especially those with fractures that included minor features. In our study, atypical fractures occurred among those with and without bisphosphonate dispensings. Seventy percent of subjects with an atypical fracture had no recorded dispensing of bisphosphonates at any time prior to the fracture event. Moreover, we found that rates of these fractures remained stable, despite increased dispensings among HMO members over the study period.21 Others have evaluated factors associated with FSF but could not separate those with atypia, due to lack of radiographic review. Abrahamsen and colleagues17 described higher rates of oral glucocorticosteroid use and similar rates of bisphosphonate dispensings in FSFs compared to hip fractures. Park-Wyllie and colleagues18 found, in a case-control study of bisphosphonate users, that, compared to transient bisphosphonate use, treatment for 5 years or more was associated with an increased risk of FSF. Our study was not designed to determine factors associated with atypical femur fractures. However, our findings suggest that if bisphosphonates are associated with atypical femur fractures, the use of this medication is one of multiple associated factors.

This study serves to create a more precise case definition of atypical femoral fractures. Individuals with atypical fractures that included only major criteria appear to be different (older and with lower BMD) from people with atypical fractures that had both major and minor criteria. These findings suggest that future epidemiologic research, especially research that focuses upon risk factors, should separately evaluate fractures with minor features. Although the numbers of atypical fractures observed are few and thus it is difficult to make definitive conclusions, the proportion of atypical cases with the minor criteria appears to be increasing over time. It is possible that the overall population-based risk of FSF has remained stable but that bisphosphonates and/or another risk factor are making it more likely for those fractures to have radiographic atypia. This concept might help explain the discrepancy between our epidemiologic results and those of other case series.

This study had several limitations. Although we observed nearly 1.25 million person-years, the condition of interest is rare and even larger populations are necessary to draw firmer conclusions about incidence rates, trends, and associated factors. Our study was conducted in one region of an HMO in a relatively homogenous population. Also, data were limited by being collected in the course of usual clinical care, as opposed to in a research setting, resulting in a quarter of the cases not having radiographs available and in BMD data not being widely available. Thus, the proportion of FSFs that were atypical may be underestimated by as much as 25%. Our study characterized the epidemiology of FSF, identifying a subset with radiographic atypia; these fractures should be the focus of future risk factor evaluation.

Our conclusion that atypical fractures of the femur have had a low and stable incidence rate from 1996 to 2009 reduces the likelihood that these fractures are solely related to bisphosphonate use. Further studies will be necessary to determine the incidence rates of atypical fractures in other geographic areas and in diverse populations. Larger studies with alternative study designs are necessary to more definitively determine which factors are associated with atypical femur fractures.

Disclosures

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References
  10. Supporting Information

ACF, NP, and AGR have received research grant funding from Merck and Amgen. DB and MMR have received research grant funding from Merck. DF has served as paid consultant to Accumed, and has received a research grant and has served as a speaker for Synthes, USA. DB has research projects funding by Novartis, Merck, and Roche. EO has industry-derived research support from Eli Lilly, Amgen, and Merck; he has consultant support from Eli Lilly, Merck, Amgen, and Wright Medical Technology. JMC and AS are employees and stockholders of Merck.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References
  10. Supporting Information

This work was supported by grants from Merck. We acknowledge the outstanding efforts of Data Coordination (Donna Gleason, Celeste Machen, and Kim Olson), Research Clinic (Ann Macfarlane, Kristin Muessig, and Judy Sanseri), and KPNW Interstate South Imaging (Russell Howard and staff) in collecting clinical data and assisting with the accuracy and completeness of the data. We also acknowledge Leslie Bienen for editorial support and Dixie Sweo for administrative assistant support.

Authors' roles: ACF directed the conception and design, and participated in the acquisition, analysis, and interpretation of data; DB participated in the conception and design, and in the analysis and interpretation of data; NP participated in the conception and design, and directed the acquisition of data and the analysis and interpretation of data; AGR and MMR participated in the conception and design, and supported NP and ACF in the acquisition, analysis, and interpretation of data; DF and DB participated in the conception and design, and in the acquisition, analysis, and interpretation of data; RD participated in the conception and design, and in the acquisition and interpretation of data; AS and JMC participated in the conception and design, and in the interpretation of data; EO participated in the conception and design and in the interpretation of data. ACF and NP accept responsibility for the integrity of the data analysis. All authors participated in drafting the manuscript or revising it critically for important intellectual content and approved the final version of the submitted manuscript.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References
  10. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References
  10. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
jbmr_1550_sm_Suppl-Radiograph-A.tif2812KRadiograph A – Atypical fracture with major features only
jbmr_1550_sm_Suppl-Radiograph-B.tif1664KRadiograph B – Atypical fracture with major and minor radiograph features pre-fracture
jbmr_1550_sm_Suppl-Radiograph-C.tif1632KRadiograph C – Atypical fracture with major and minor radiograph features post-fracture
jbmr_1550_sm_Supplmental-1.doc24KSupplemental 1: ICD-9 Fracture Code Categories
jbmr_1550_sm_Supplmental-2.doc36KSupplemental 2: FSF* missing radiographs by year
jbmr_1550_sm_Supplmental-3.doc34KSupplemental 3: Changes in FSF* fracture category at various levels of clinical review

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.