Morphology of the Proximal Femur Differs Widely with Age and Sex:Relevance to Design and Selection of Femoral Prostheses

David S. Casper,1 Gregory K. Kim,1 Javad Parvizi,1 Theresa A. Freeman2

1The Rothman Institute of Orthopaedics at Thomas Jefferson University Hospital, 925 Chestnut Street, 5th Floor, Philadelphia, Pennsylvania19107, 2Thomas Jefferson University, Department of Orthopaedic Surgery, 1015 Walnut Street, Suite 501, Philadelphia, Pennsylvania 19107

Received 24 September 2011; accepted 5 December 2011

Published online in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jor.22052

ABSTRACT: The ability of uncemented femoral stems to osseointegrate properly depends largely on their fit in the proximal femur. Weevaluated the topography of the proximal femur and determined differences based on age and sex. Retrospectively, anteroposteriorradiographs from 312 (168 male, 144 female) pre-operative total hip arthroplasty (THA) patients (age of 21–85 years) were collected.Radiographic measurements were taken at 10 mm intervals along the length of the femur. Variables including canal flare index (CFI)and cortical index (CI) were calculated. Data were binned into three age groups and separated by sex for comparison. Measurementsshowed that CFI decreased with age for both sexes; however, females demonstrated a greater decrease. Decrease in flare occurredprimarily on the lateral side. CI also decreased with age, the most pronounced drop occurring in older females. A clear difference existsbetween male and female proximal femoral geometry. This decrease is most likely attributed to the loss of cortical bone. The medialcomponent likely demonstrates less loss of flare due to strong compressive forces that are transmitted through this portion of the femur.These results demonstrate the necessity of considering age and sex when selecting a proper prosthesis. � 2012 Orthopaedic ResearchSociety. Published by Wiley Periodicals, Inc. J Orthop Res 9999:1–5, 2012

Keywords: joint arthroplasty; proximal femur; canal flare

Total hip arthroplasty (THA) is among the most suc-cessful surgical procedures available today.1–3 With itscontinued success, THA is increasingly being offeredto younger patients.4 One of the predictors of THA out-come is appropriate positioning of the components. Inthe case of uncemented THA, it is imperative that theprosthetic stem fit correctly in the proximal femur.5–10

Proper fit assures bony ingrowth into the prosthesis,resulting in long-term fixation and stability.

Preoperative templating is performed in clinicalpractice to determine the size of femoral stem that islikely to have the best fit. Further intraoperativerefinements are made to select an appropriate stem.Despite these efforts, undersizing of the stem, leadingto lack of osseointegration, or oversizing, leading toperiproshetic fracture, do occur.11–13

It is unknown what other factors, if any, could beused to ensure appropriate sizing and fit of femoralcomponents. Previous investigators devoted effort tounderstand the morphology of the proximal femur andidentify factors that may guide surgeons in implant se-lection. The CFI is one such metric; it is determinedby measuring the diameter of the proximal femur20 mm proximal to the lesser trochanter and at theisthmus to generate a ratio.5 Elderly patients withthin diaphyseal cortices are expected to have a smallerCFI compared to younger patients.6

Additionally, several studies indicated initial dif-ferences in femoral canal dimensions between sexesand other differences were also noted as occurringwith age.6,14 Noble et al. used radiographic measure-ments from 200 cadaveric specimens from 22 to95 years of age. In a follow-up study, where old and

young males and females were classified; the CFI ofmen exhibited little change with age, whereas the CFIfor women showed a large decrease.6 The investigatorsattributed this finding to the loss of cortical bone withage.6,14

Other investigators used computerized tomography(CT) images to map the proximal femur and concludedthat differing CFIs were influenced by the mean age ofthe patients sampled.14–16 Although CT scans weremost accurate in these studies, the resultant CFImeasurements did not differ greatly from a 2D radio-graphic analysis.16 As THA patients are most oftenevaluated by plain radiographs preoperatively andagain for routine postoperative follow-ups, their avail-ability makes them the most practical and logicalchoice in clinical use.

We hypothesized that various radiographic param-eters can be determined for use in clinically selectingthe most appropriate femoral implant for a givenpatient. Our study is unique in several aspects, as wehave included a large cohort of patients between 20and 40 years of age, and we used semi-automatedmeasurements from digital radiographs using softwaredeveloped at our institution. We also included adetailed analysis comparing age and sex in regards tocortical index (CI), investigated the relationshipbetween CI and CFI, and offered a more detailed anal-ysis of CFI changes with sex and age. Additionally, theCFI changes were dissected to determine the individualcontribution of the lateral versus medial flare.

MATERIALS AND METHODSPatient PopulationWe utilized a standardized retrieval protocol of preoperativeserial AP radiographs from 168 male and 144 femalepatients, aged 21–85, who underwent THA between 2008and 2009 at our institution; patients were selected based

Correspondence to: Theresa A. Freeman (T: þ215-955-1068;F: þ215-955-9159; E-mail: Theresa. Freeman@jefferson.edu)

� 2012 Orthopaedic Research Society. Published by Wiley Periodicals, Inc.

JOURNAL OF ORTHOPAEDIC RESEARCH 2012 1NL12-AR-HP-334

on age and sex. These patients primarily reside in the north-eastern United States and come predominately from a Cau-casian background. Detailed demographic data werecollected that included age, sex, Body Mass Index (BMI), andother variables (Table 1). Data collection was in accord withIRB guidelines of Thomas Jefferson University Hospital.Radiographs from patients with prior trauma involving thefemur were excluded.

Data CollectionHigh-resolution digital radiographs were used for all analy-ses. Current practice in digital radiography is the inclusionof a magnification scale on all acquired radiographs, andpatients are aligned in a standard manner to produce mini-mal rotation of the hip during image collection. These proce-dures help standardize radiographs among patients. Toconfirm accuracy and reproducibility of the measurementsoftware, radiographs of the same hip imaged on separatevisits were evaluated. This comparative analysis was per-formed on four patients, and the measurement differencewas <2 mm, only slightly above the standard deviationobtained by measuring the same radiograph over again. Todetermine reproducibility, the same radiograph was mea-sured by several individuals, producing a standard deviationof <1 mm. Additionally, to determine intra-observer variabil-ity, the same radiograph was measured several times by asingle user, also producing a standard deviation of <1 mm.

Using a custom programmed, semi-automated analysissoftware module within ImagePro Plus (Mediacybernetics,Silver Springs, MD), the AP view of the pelvis was used tomeasure femoral canal dimensions after a single investigatormanually selected three points: (i) the midpoint at the isth-mus; (ii) the most medial aspect of the lesser trochanter; and(iii) the midpoint of the proximal metaphysis. The programthen produced a circle, which the user centered over the fem-oral head. Using these inputs, the program automaticallycreated an axis along the length of the canal, with measure-ments at 10 mm intervals along this axis beginning 20 mmproximal to the lesser trochanter and ending 110 mm distally(Fig. 1). An automated edge detection algorithm was appliedbased on the difference in pixel density detected on the linedrawn out from the center point. In this way, a consistentdistance for the canal border was determined based on radio-graphic density in a non-subjective manner. In the case ofpoor radiographic quality at any given point, a manual over-ride was performed. All values were then exported to a pre-formatted spreadsheet. A single observer was used to collectthe data to minimize user variability.

Data AnalysisThe values generated by the above analysis were used to cal-culate the CI (cortical bone at the isthmus/periosteal widthat the isthmus) and the CFI (canal width 20 mm proximal tothe lesser trochanter/endosteal width at the isthmus). TheCFI was then subdivided into the lateral flare index (LFI)

and medial flare index (MFI). These ratios were calculated inthe same fashion as the CFI, using only the medial or lateralcomponents of the total canal width.

After radiographic data collection was completed, datawere separated based on sex and age. The age brackets wereyoung (20–40), middle aged (41–60), and elderly (61–85). Thefemale age ranges (mean � SD) were 22–40 (34 � 5.6), 41–60 (50 � 4.12), and 61–85 (74 � 7.63), with 25, 88, and 31patients in each group, respectively. The male age rangeswere 21–40 (34 � 5.27), 41–60 (49 � 3.82), and 61–84(71 � 6.17), with 29, 115, and 24 patients in each group.

In an effort to determine what changes occur in theproximal femur with age and sex, the mean and SD of eachmeasurement was calculated and then recalculated afterseparation by sex and then by age and sex.

Statistical AnalysisStatistical analysis between each cohort was performed usinga two-tailed Student’s t-test, and a p-value <0.05 was used todetermine level of significance (a).

RESULTS

Classification of Proximal Femur Dimensions by AgeOverall data were first binned by age. The mean values(with SD) for each collected statistic were then calcu-lated (Table 2). These values demonstrated an overall

Table 1. Patient Demographic Data

N Age (Mean) BMI (Mean) Right Left

Total 312 50.7 � 12.6 28.4 � 5.3 159 153Male 168 49.6 � 11.2 29.7 � 4.9 91 77Female 144 52.0 � 13.9 27 � 5.5 68 76

Figure 1. Macro generated measurements of the proximalfemur. Measurements begin 20 mm proximal to the lesser tro-chanter and proceed distal via 10 mm increments until reachinga point 110 mm distal to the lesser trochanter.

2 CASPER ET AL.

JOURNAL OF ORTHOPAEDIC RESEARCH 2012NL12-AR-HP-334

decrease in CFI with age, (4.0 young, 3.9 middle aged,and 3.5 elderly). This decrease was noted both medial-ly and laterally; however, the lateral componentshowed the greatest decrease with age (4.2 young, 4.1middle aged, and 3.4 elderly). CI also showed a decreas-ing trend with age, with elderly patients demonstratingthe lowest ratio (0.62 young, 0.61 middle aged, and 0.58elderly). Analyses of these data were conducted to deter-mine values of significance (Table 2). No significancewas found when comparing the young and middle-aged patients for any variable other than BMI. MFIwas not significant. Significant values were calculatedwhen comparing the young and elderly, as well as themiddle aged and elderly for CI, CFI, and LFI. BMIwas also significant when comparing middle agedpatients with the elderly.

Classification of Proximal Femur Dimensions byAge and SexThe data were then separated by sex, and again sortedby age. Overall, male and female data demonstratedsimilar trends to that of the overall population(Tables 3 and 4). Females demonstrated a largedecrease in LFI (4.6 young, 4.4 middle aged, and 3.4elderly) with age compared to a more gradual decreaseof the MFI (4.2 young, 3.8 middle aged, and 3.6 elder-ly). Significance for females was seen when comparingthe young and elderly and the middle aged and elderlyfor CI, CFI, and LFI. For females, MFI demonstratedsignificance when comparing the young and elderlypatients; BMI showed for no significant (Table 3).Males had similar trends, but to a lesser degree, withthe MFI producing a fairly constant value throughoutthe age groups (3.5 young, 3.6 middle aged, and 3.6elderly) (Table 4). Few significant values were notedfor males (Table 4). The significant values are LFI

when comparing young and elderly and middle agedand elderly, and BMI when comparing young and mid-dle aged and middle aged and elderly.

Comparison of Sex DataOf note, when comparing sexes, females demonstrateda greater initial flare in the young for both MFI(female 4.2, male 3.5) and LFI (female 4.6, male 4.0).For CFI, young females had a mean value of 4.4, whilethe young males’ mean value was 3.8. Elderly patientsdemonstrated the same CFI value of 3.5.

When comparing age groups of different sexes, afew variables arose as significant (Table 5), including

Table 2. Mean Values (and SD) of Patient Statistics Ac-companied by p-values Providing Comparisons of AgeGroups, significant values are bolded (p < 0.05)

Overall n ¼ 54 n ¼ 203 n ¼ 55

Age 21–40 41–60 61–85BMI 27.3 � 5.31 29 � 5.49 26.8 � 4.18CorticalIndex

0.62 � 0.07 0.61 � 0.05 0.58 � 0.06

CFI 4 � 0.91 3.9 � 0.78 3.5 � 0.54LFI 4.2 � 1.1 4.1 � 1.01 3.4 � 0.7MFI 3.8 � 0.92 3.7 � 0.83 3.6 � 0.68Age 37 � 2.29 49 � 3.94 73 � 7.17

20–40 vs.41–60

20–40 vs.61–85

41–60 vs.81–85

CorticalIndex

0.1217 0.0004 0.0003

CFI 0.1401 0.0000 0.0002LFI 0.2384 0.0000 0.0000MFI 0.1855 0.0615 0.3979BMI 0.0463 0.5711 0.0060

Table 3. Mean Values (and SD) of Female Patient Sta-tistics Accompanied by p-values Providing Comparisonsof Age Groups, significant values are bolded (p < 0.05)

Female n ¼ 25 n ¼ 88 n ¼ 31

Age 22–40 41–60 61–85BMI 26.4 � 5.71 27.2 � 5.64 26.9 � 5.1Cortical

Index0.62 � 0.05 0.61 � 0.05 0.56 � 0.06

CFI 4.4 � 0.72 4.1 � 0.81 3.5 � 0.51LFI 4.6 � 0.88 4.4 � 1.04 3.4 � 0.68MFI 4.2 � 0.84 3.8 � 0.86 3.6 � 0.65Age 34 � 5.6 50 � 4.12 74 � 7.63

20–40 vs.41–60

20–40 vs.61–85

41–60 vs.81–85

CorticalIndex

0.1798 0.0002 0.0000

CFI 0.0762 0.0000 0.0002LFI 0.3228 0.0000 0.0000MFI 0.0348 0.0038 0.2769BMI 0.5497 0.7568 0.7750

Table 4. Mean Values (and SD) of Male Patient Statis-tics Accompanied by p-values Providing Comparisons ofAge Groups, significant values are bolded (p < 0.05)

MALE n ¼ 29 n ¼ 115 n ¼ 24

AGE 21–40 41–60 61–84BMI 28.1 � 4.84 30.4 � 4.98 26.8 � 2.66Cortical

Index0.62 � 0.08 0.61 � 0.06 0.6 � 0.05

CFI 3.8 � 0.62 3.7 � 0.64 3.5 � 0.49LFI 4 � 3.98 3.7 � 3.75 3.5 � 3.48MFI 3.5 � 3.55 3.6 � 3.62 3.6 � 3.59Age 34 � 5.27 49 � 3.82 71 � 6.17

20–40 vs.41–60

20–40 vs.61–85

41–60 vs.81–85

CorticalIndex

0.3545 0.2193 0.3858

CFI 0.9049 0.0758 0.0553LFI 0.5749 0.0081 0.0070MFI 0.6417 0.8230 0.8504BMI 0.0291 0.2284 0.0007

VARIATION IN THE PROXIMAL FEMUR 3

JOURNAL OF ORTHOPAEDIC RESEARCH 2012NL12-AR-HP-334

all flare indices for young patients, CFI, LFI, and BMIfor middle-aged patients, but only CI for the elderlypatients.

DISCUSSIONThe geometry of the proximal femur has long beenknown as critical in the design of femoral stems forTHA.5–9 The recent trend for performing THA surger-ies on younger patients raises the question as to if thesame design of component can be used in all agegroups and if a difference exists in the morphology ofproximal femur in different age groups and sexes.6 Weinvestigated key indices of femoral geometry withrespect to age and sex, including a large cohort of 20-to 40-year-old THA patients, in addition to the moretypical 41–60 and 61–80 age groups.

Overall, our results were similar to those of Nobleet al.5 and Husman et al.14 The main, and perhapsmost important, difference between our study and pre-vious reports is that a larger number of young patientswere included in our study. The average age of ourpopulation was 51, producing a canal flare of 3.85(1.84–6.80), while Noble et al.5 had a mean age of 70and a mean canal flare of 3.8 (2.4–7.0), and Husmanet al.14 had a mean age of 62 and a mean canal flare of3.81 (2.3–7.2). The inclusion of young patients is im-portant, as recently concern has arisen with the use oftraditional proximally coated uncemented femoralcomponents in younger patients.17,18 Because of thehigher density of diaphyseal cortices and a high flare,conventional femoral stems are likely to ‘‘hang up’’ inthe diaphyses, leading to thigh pain and lack ofosseointegration.19 Our study suggests that youngpatients do indeed have a high CFI and CI and mayrequire a different type of stem design aimed at reduc-ing the diameter of the stem in the diaphyseal regionfor a given stem size.

We also found that CFI and CI decreased with age,consistent with previous studies.6,14 We found that theloss of cortical bone is most readily seen when compar-ing middle aged with elderly patients (middle aged0.61, elderly 0.58). A minimal decrease was foundwhen comparing young patients with the middle aged(young 0.62, middle aged 0.61). This loss of corticalbone is the primary reason for the decrease in canalflare. A small decrease in the cortices leads to anincrease in isthmus width, which ultimately leads to adecrease in the overall flare.

When the data were segmented by both age andsex, the 20–40 age canal geometry had a significantlylarger flare than males, with an initial flare of 4.4decreasing to 3.5 in the 61–80 cohort, compared to themale 20–40 cohort with a mean of 3.8 decreasing mod-estly to 3.5. One can speculate that the cause of thisinitial CFI difference in women may be related to dif-ferences in pelvic shape, weight distribution, childbirth, and characteristically wider hips. However,smaller stature may also play a significant role, asAsian populations, such as Koreans, are also reportedto have larger CFI measurements without regardto sex.15 To date, these differences have not musteredmuch clinical concern. However, the significant dif-ferences in MFI between males and females of theyoungest cohort may cause more concern as thenumber of THA increases in this population. Youngermales apparently exhibit fewer significant differencesin their canal parameters compared with oldercohorts, whereas younger females not only show signif-icant differences to younger males, but also to olderfemale.

To determine where in the canal this flare occurs, amore in depth analysis was performed by dividing themeasurement into medial and lateral components. Bydifferentiating the canal flare, we were better able todetermine where the decrease in overall flare occurredwith age. Our radiographic measurements demon-strated that the decrease primarily occurs on the later-al side for both males and females. This is likelyrelated to the loss of cortical bone on the lateral por-tion of the proximal femur. The lower loss of corticalbone along the medial portion of the femur is thoughtto be due to the presence of strong compressive forcesfrom load transfer through the skeletal structure. Asthe weight is transmitted through the femur, the pri-mary compressive forces promote bone growth and sta-bilization along the medial side. The lateral side isresponsible for supporting the secondary tensile forces,and thus it is more likely that the cortical bone that islost with age is primarily lost from the lateral sidewhere less load is transmitted.

Another possibility as to why the lateral componentis responsible for the decrease in flare is related to themeasurement itself. The lateral measurement proxi-mal to the lesser trochanter often falls on an area inthe proximal metaphysis that lacks cortical bone.Therefore, as a patient ages, no bone exists to be lostat this measurement point, while cortical bone is lostat the distal measurement resulting in a decreasedflare. Medially, cortical bone is present both at themeasurement point proximal to the lesser trochanterand at the isthmus. This leads to a greater width atboth measurement points, which results in a smallerdecrease.

Our study has limitations, the first of whichincludes user bias in the determination of the canalboundary. To help eliminate different biases in termsof selection of corrected points along the boundary, one

Table 5. p-values of Sex Comparisons Among AgeGroups, significant values are bolded (p < 0.05)

Male vs. Female 21–40 41–60 61–85

Cortical Index 0.7669 0.7370 0.0374CFI 0.0008 0.0009 0.8688LFI 0.0081 0.0001 0.8048MFI 0.0022 0.1652 0.9972BMI 0.2423 0.0000 0.9458

4 CASPER ET AL.

JOURNAL OF ORTHOPAEDIC RESEARCH 2012NL12-AR-HP-334

user was responsible for all measurements, and azoom feature of the software was utilized to clearly de-lineate the boundary of cortical bone. Another sourceof error includes the use of radiographic images. Asmall amount of rotation of the hip could have affectedmeasurement reproducibility, but measurement of thesame hip at different acquisition times produced onlyminimal differences, and with such a large cohort thiseffect is thought to be minimal. Radiographs areknown to be less accurate than CT scans; however, itis believed that the radiographs produced adequatemeasurements to establish an overall trend. Lastly,the isthmus was taken to be at 110 mm for all thepatients. While studies showed that this is an averagelocation, its position is variable. This may haveresulted in differing values for the flare indices, al-though the results produced were similar to thoseshown in previous radiographic studies.

Proximal femur geometry is important for bothchoosing an uncemented implant for a given patientand for the design of these implants. Several studiesdocumented this geometry; however, few included alarge cohort of young patients and the separation ofboth sex and flare components. By better understand-ing how sex and age relate to the attributes of theproximal femur, surgeons can more likely choose animplant that will properly osseointegrate and achieveexcellent fixation. Based on our findings, patients withwide canal (reduced CI and high CFI), especially theelderly, should probably receive a cemented femoralstem. Patients in the opposite end of the spectrumwith type A bone, high CI, and low CFI, should also beevaluated carefully, and if necessary be reconstructedwith a modular stem to prevent diaphyseal hang-upand subsequent failure. In recent years, having recog-nized the limitation of tapered metaphyseal stems,manufacturers are investing efforts to produce shorterstems and stems with non-proportional metaphyseal todiaphyseal ratios. Our results demonstrate the neces-sity of considering age and sex when selecting theproper prosthesis.

ACKNOWLEDGMENTSThe authors would like to thank Mark Sobiesk for his helpprogramming the module used for the analysis in ImagePro. This work was completed with funding from StrykerOrthopaedics (Mahwah, NJ) and The Rothman Institute(Philadelphia, PA). J. Parvizi is a consultant for StrykerOrthopaedics (Mahwah, NJ) and has intellectual property onSmarTech (Philadelphia, PA).

REFERENCES1. McLaughlin JR, Lee KR. 1997. Total hip arthroplasty with

an uncemented femoral component. Excellent results at ten-year follow-up. J Bone Joint Surg 79-B:900.

2. Parvizi J, Keisu KS, Hozack WJ, et al. 2004. Primary totalhip arthroplasty with an uncemented femoral component: along-term study of the Taperloc stem. J Arthroplasty 19:151–156.

3. McLaughlin JR, Lee KR. 2008. Total hip arthroplasty withan uncemented femoral component. J Bone Joint Surg 90:1290–1296.

4. Kurtz SM, Lau E, Ong K, et al. 2009. Future young patientdemand for primary and revision joint replacement. ClinOrthop Relat Res 467:2606–2612.

5. Noble PC, Alexander JW, Lindahl LJ, et al. 1988. The ana-tomic basis of femoral component design. Clin Orthop RelatRes 235:148–165.

6. Noble PC, Box GG, Kamaric E, et al. 1995. The effect ofaging on the shape of the proximal femur. Clin Orthop RelatRes 316:31–44.

7. Dorr LD, Faugere MC, Mackel AM, et al. 1993. Structuraland cellular assessment of bone quality of the proximalfemur. Bone 14:231–242.

8. Fessy MH, Seutin B, Bejui J. 1997. Anatomical basis for thechoice of the femoral implant in the total hip arthroplasty.Surg Radiol Anat 19:283–286.

9. Laine HJ, Lehto MU, Moilanen T. 2000. Diversity of proxi-mal femoral medullary canal. J Arthroplasty 15:86–92.

10. Lecerf G, Fessy MH, Philippot R, et al. 2009. Femoral offset:anatomical concept, definition, assessment, implications forpreoperative templating and hip arthroplasty. OrthopTraumatol Surg Res 95:210–219.

11. Soballe K, Hansen ES, B-Rasmussen H, et al. 1992. Tissueingrowth into titanium and hydroxyapatite-coated implantsduring stable and unstable mechanical conditions. J OrthopRes 10:285–299.

12. Kroon PO, Freeman MA. 1992. Hydroxyapatite coating ofhip prostheses: effect on migration into the femur. J BoneJoint Surg Br 74:518–522.

13. Berry DJ. 2002. Management of periprosthetic fractures: thehip. J Arthroplasty 17:11–13.

14. Husmann O, Rubin PJ, Leyvraz PF, et al. 1997. Three-dimensional morphology of the proximal femur. J Arthro-plasty 12:444–450.

15. Khang G, Choi K, Kim CS, et al. 2003. A study of Koreanfemoral geometry. Clin Orthop Relat Res 406:116–122.

16. Rubin PJ, Leyvraz PF, Aubaniac JM, et al. 1992. Themorphology of the proximal femur. A three-dimensionalradiographic analysis. J Bone Joint Surg Br 74:28–32.

17. Liang TJ, You MZ, Xing PF, et al. 2010. Uncemented totalhip arthroplasty in patients younger than 50 years: a 6- to10-year follow-up study. Orthopedics 16:236–239.

18. McLaughlin JR, Lee KR. 2011. Total hip arthroplasty withan uncemented tapered femoral component in patients youn-ger than 50 years. J Arthroplasty 26:9–15.

19. Kinov P, Radl R, Zacherl M, et al. 2007. Correlation betweenthigh pain and radiological findings with a proximallyporous-coated stem. Acta Orthop Belg 73:618–624.

VARIATION IN THE PROXIMAL FEMUR 5

JOURNAL OF ORTHOPAEDIC RESEARCH 2012NL12-AR-HP-334