G Model

JINJ-4780; No. of Pages 8

Injury, Int. J. Care Injured xxx (2011) xxx–xxx

Contents lists available at SciVerse ScienceDirect

Injury
journal homepage: www.elsevier.com/locate/injury

Computer navigation in the reduction and fixation of femoral shaft fractures:

A randomized control study§
Oliver Keast-Butler a, Michael J. Lutz b, Mark Angelini c, Nick Lash d, Dawn Pearce e,
Meghan Crookshank f, Rad Zdero f,g,*, Emil H. Schemitsch f,h
a
Conquest Hospital, Hastings, United Kingdom
b
Royal Brisbane Hospital, Herston, Queensland, Australia
c
Etobicoke General Hospital, Toronto, ON, Canada
d
Hawkes Bay Hospital, New Zealand
e
Medical Imaging, St. Michael’s Hospital, Toronto, ON, Canada
f
Martin Orthopaedic Biomechanics Laboratory, St. Michael’s Hospital, Toronto, ON, Canada
g
Dept. of Mechanical and Industrial Engineering, Ryerson University, Toronto, ON, Canada
h
Faculty of Medicine, Department of Surgery, University of Toronto, Toronto, ON, Canada

A R T I C L E I N F O A B S T R A C T

Article history: Objectives: We investigated the accuracy of reduction of intramedullary nailed femoral shaft fractures in
Accepted 17 August 2011 human cadavers, comparing conventional and computer navigation techniques.
Methods: Twenty femoral shaft fractures were created in human cadavers, with segmental defects
Keywords: ranging from 9 to 53 mm in length (Winquist 3–4, AO 32C2). All fractures were fixed with antegrade
Randomized control study 9 mm diameter femoral nails on a radiolucent operating table. Five fractures (‘‘Fluoro’’ group) were fixed
Computer navigation with conventional techniques and fifteen fractures (‘‘Nav 1’’ and ‘‘Nav 2’’ groups) with computer
Femur shaft fracture
navigation, using fluoroscopic images of the normal femur to correct for length and rotation.
Reduction
Fixation
Postoperative CT scans compared femoral length and rotation with the normal leg.
Results: Mean leg length discrepancy in the computer navigation groups was smaller, namely, 3.6 mm
for Nav 1 (95% CI: 1.072 to 6.128) and 4.2 mm for Nav 2 (95% CI: 0.63 to 7.75) vs. 9.8 mm for Fluoro (95%
CI: 6.225 to 13.37) (p < 0.023). Mean rotational discrepancies were 8.78 for Nav 1 (95% CI: 4.282 to 13.12)
and 5.68 for Nav 2 (95% CI: 0.65 to 11.85) vs. 9.08 for Fluoro (95% CI: 2.752 to 15.25) (p = 0.650).
Conclusions: Computer navigation significantly improves the accuracy of femoral shaft fracture fixation
with regard to leg length, but not rotational deformity.
ß 2011 Elsevier Ltd. All rights reserved.

Introduction The ‘gold standard’ treatment for treating these fractures is

Femoral shaft fractures are frequently encountered in ortho-
paedic surgery. The epidemiology of femoral shaft fractures has
been studied in various countries. One study reported 37.1
incidences per 100 000 person years for the period 1965–1984.1
Another investigation tracked 9.9 fractures per 100 000 person
years between 1985 and 1994, with peak incidences seen in males
between 15 and 24 years of age and in females over the age of 75
years, occurring as a result of high-energy trauma and osteoporo-
sis, respectively.2
§
Devices and materials: The focus of this investigation was assessment of
rotational and leg length deformity following reduction and antegrade nailing of
femoral shaft fractures using computer navigation and conventional techniques.
* Corresponding author at: Biomechanics Lab – St. Michael’s Hospital, Li Ka Shing
Building, West Basement, Room B114/B116, 38 Shuter St, Toronto, ON, Canada M5B
1W8. Tel.: +1 416 864 5482; fax: +1 416 359 1601.
E-mail address: zderor@smh.ca (R. Zdero).
fixation with intramedullary nails. The original technique of
unlocked nailing has a high rate of union with a low incidence of
complications.3 By using locking screws proximally and distally,
improved control of length and rotation is achieved, extending the
indications of femoral nailing to proximal and distal fractures and
to fractures with unstable fracture patterns.4
Malunion is a possible complication following any fracture
fixation. In femoral shaft fractures, length, angulation and rotation
must be accurately corrected during the intramedullary nailing
procedure. Reported incidence of postoperative shortening and
malrotation varies. In a series of 520 femoral fractures,4 shortening
of >1 cm occurred in 9% of cases, and malrotation of >108 occurred
in 10% of cases. In a recent prospective randomised study, the
incidence of shortening >1 cm or malrotation >108 were 7% and
34%, respectively, in the fracture table treatment group.5
Computer navigation has evolved as a tool in orthopaedic
surgery. With the use of a camera and trackers on patients, the
computer is able to monitor the position and orientation of surgical

0020–1383/$ – see front matter ß 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.injury.2011.08.020

Please cite this article in press as: Keast-Butler O, et al. Computer navigation in the reduction and fixation of femoral shaft fractures: A
randomized control study. Injury (2011), doi:10.1016/j.injury.2011.08.020
G Model
JINJ-4780; No. of Pages 8
2 O. Keast-Butler et al. / Injury, Int. J. Care Injured xxx (2011) xxx–xxx
instruments and bones. Image-free systems use predetermined
anatomical models and combine these with intra-operative
information to customise the model to the specific patient.
Image-based systems combine intra-operative information with
radiological images. These images can be acquired either
preoperatively, usually in the form of computed tomography
(CT) or magnetic resonance (MR) scans, or intra-operatively, with
the use of a modified fluoroscopic C-arm. However, no studies to
date have investigated the use of this technology specifically for
improved reduction of femoral shaft injury.
The aim of this study was to investigate whether a novel
computer-assisted technique for reducing femoral shaft fractures
would improve the accuracy of fracture fixation in a human
cadaveric model of femoral shaft fractures compared with
conventional reduction techniques.
Methods
General approach
Eighteen matched-pair cadaveric femurs were used, two of
which were used twice (Table 1). Right femurs underwent fracture
and repair using one of three methods. A ‘Fluoro’ control group
(n = 5) was repaired using traditional fluoroscopic techniques. A
‘Nav 1’ (n = 10) and a modified ‘Nav 2’ (n = 5) study group were
repaired employing computer navigation. All left femurs remained
intact. Mean leg length discrepancy and rotational deformity from
the three repair groups were finally compared with contralateral
intact femurs. The same experienced fellow performed all
surgeries with the assistance of co-workers.
Specimen preparation
Twenty full body human cadavers embalmed with introfiant
were obtained with permission of our institution’s ethics board.
The cadavers were examined visually and fluoroscopically to
confirm they had undergone no previous surgery involving either
femur. Two cadavers were identified as having had such surgery
and were excluded. Therefore, two of the remaining cadavers were
used twice; the difficulty in obtaining two new cadavers and the
financial costs involved made this decision necessary.
Fractures in the diaphysis were created to model an unstable
fracture (Winquist 3/4, Arbeitsgemeinschaft für Osteosynthese-
fragen (AO) 32-C2 classification). Using an anterior approach with
splitting of the quadriceps muscle, sections of bone ranging from 5
to 50 mm in length were excised from the femoral diaphysis and
removed from the thigh. The fracture area was completely stripped
of soft tissue for a distance of 10 cm proximally and 10 cm distally,
simulating the soft-tissue injury that occurs in a femoral
diaphyseal fracture. Prior to completion of the osteotomies, the
defect length was measured with a Vernier calliper, having a
precision of 0.1 mm. The measured defect included the length of
resected bone and the bone removed by the thickness of the saw
blade for each of the two cuts made. The wound was then closed,
making the fracture representative of a closed femoral shaft
fracture. All left femurs remained intact.
Surgical repair technique
All fractures were fixed using 9 mm diameter antegrade
femoral nails on a radiolucent operating table. The proximal
femur was reamed to a 12-mm diameter to accommodate the
chosen design of femoral nail, and the diaphysis was reamed to a
10.5-mm diameter prior to nail insertion. The nail was locked with
proximal and distal locking bolts. It should be noted that the real
advantage of locking bolts was that they helped maintain whatever
Please cite this article in press as: Keast-Butler O, et al. Computer navigation in the reduction and fixation of femoral shaft fractures: A
randomized control study. Injury (2011), doi:10.1016/j.injury.2011.08.020
rotational reduction was achieved surgically with the femoral nails
(whether excellent or poor), rather than expecting the bolts to
alleviate rotational abnormalities, as such.
Fractures in the Fluoro control group (n = 5) were fixed with
conventional reduction techniques. Fluoroscopic images of the
fracture and control limb were compared to optimise length and
rotation, prior to locking of the nail. A radio-opaque ruler was used
to measure the length of normal and fractured femurs to guide
correction of length. Radiographs of both femurs were compared,
examining profiles of the proximal femur, particularly the lesser
trochanter and the knee, ensuring rotation was similar in both
femurs.6
Fractures in the Nav 1 (n = 10) and Nav 2 (n = 5) study groups
were fixed with computer navigation, using fluoroscopic images of
the normal femur to correct for length and rotation (Figs. 1 and 2).
The navigation system used active infrared trackers. The C-arm had
a tracker ring attached to the camera, which corrected image
distortion and relayed information regarding the position of the C-
arm to the computer. Trackers were fixed to the proximal and
distal femur of the fractured limb and to the midshaft of the control
limb with unicortical pins. Antero-posterior and lateral images
were obtained of the hip and knee of both femurs and of the
fracture site.
The normal left limb in each cadaver was used as the ‘template’
to guide reduction of the fractured right limb in both the computer
navigation (i.e., Nav 1 and Nav 2) and control (i.e., Fluoro) groups.
Consequently, accurate reduction was intended to give equal limb
length and alignment in both groups.
Computer navigation for Nav 1 and Nav 2 groups
The iNstride System (Stryker, Mahwah, NJ, USA) was used in
this study. Using the images of the hips, the centre of the femoral
head, the piriform fossa and the proximal medullary canal (at the
level of the lesser trochanter) were identified on both antero-
posterior and lateral views. Using the images of the knees, the
distal joint line was recorded on the antero-posterior view, and the
centre of the epicondyles (defined as the centre of a sphere, placed
on the distal femoral joint line) was recorded on the lateral view
(Fig. 3). The computer created intersecting planes from these
landmarks, allowing calculation of the femoral neck anteversion,
the trans-epicondylar axis and the femoral length, in relation to the
control femur.
By digitising the edges of the proximal and distal fracture
fragments, fracture reduction and passage of the guide wire were
performed with virtual fluoroscopy. This allowed real-time
movement of the fracture fragments in relation to each other to
be seen without using fluoroscopy. Manual traction was applied
and, once reduction was reasonable, the distal locking screw was
inserted. Fine adjustments to length and rotation were then made
using the navigated measurements displayed on the computer
prior to proximal locking.
Two cadavers (cases 12/16 and cases 11/17) were used twice
with the two navigation software protocols to compensate for
exclusion of two previously operated specimens. To prevent bias,
the second case with each cadaver used a different length nail
locked distally with antero-posterior bolts, as opposed to lateral
bolts in the first nailing procedure.
The first 10 navigated fractures used the initial landmark
protocol described above (Nav 1). The remaining five navigated
fractures were fixed using a modified landmark protocol with an
adjusted software program (Nav 2). This adjusted protocol was
used to more reproducibly identify the piriform fossa using antero-
posterior lateral images and anteromedial/anterolateral oblique
images. Specifically, instead of using the piriform fossa as a
proximal landmark, the centre of the medullary canal at the level of
 JINJ-4780; No. of Pages 8
G Model
randomized control study. Injury (2011), doi:10.1016/j.injury.2011.08.020
Please cite this article in press as: Keast-Butler O, et al. Computer navigation in the reduction and fixation of femoral shaft fractures: A

Table 1
The characteristics of Fluoro, Nav 1, and Nav 2 groups in this investigation. A total of 20 cases were used in the Fluoro (n = 5), Nav 1 (n = 10), and Nav 2 (n = 5) groups. Fluoro results were based only on CT scans. Nav 1 and Nav 2
specimens were fixed using computer navigation and then ‘‘double-checked’’ with CT scans post surgery. Length (or rotational) discrepancy = absolute difference of fixed leg value normal leg value. Negative values ( ) for
navigation measurements indicate opposite sense compared to CT measurements, i.e. shorter vs. longer length or internal vs. external rotation. Navigation error = CT measurement navigation measurement. Key: n/a = not
applicable; * = missing data.

Case No. Group Sex Age Defect CT length discrepancy Navigation length Navigation length CT rotational Navigation rotational Navigation rotational

O. Keast-Butler et al. / Injury, Int. J. Care Injured xxx (2011) xxx–xxx

size (mm) (fixed vs. normal discrepancy (fixed vs. error (navigation vs. discrepancy (fixed vs. discrepancy (fixed error (navigation vs.
leg) (mm) normal leg) (mm) CT) (mm) normal leg) (8) vs. normal leg) (8) CT) (8)

1 Fluoro M 83 12.0 11 n/a n/a 6 n/a n/a

2 Fluoro M 86 20.5 13 n/a n/a 22 n/a n/a
3 Fluoro M 84 33.4 17 n/a n/a 0 n/a n/a
4 Fluoro M 75 52.9 8 n/a n/a 4 n/a n/a
5 Fluoro M 77 11.3 0 n/a n/a 13 n/a n/a
6 Nav 1 M 74 9.0 3 3 0 14 14 0
7 Nav 1 M 82 12.2 8 * * 17 * *
8 Nav 1 M 76 19.5 2 1 3 20 20 0
9 Nav 1 F 69 32.9 3 2 1 2 2 0
10 Nav 1 F 94 53.0 1 2 1 11 12 1
11 Nav 1 F 94 16.5 4 5 1 1 6 5
12 Nav 1 F 68 40.1 4 3 1 9 14 5
13 Nav 1 F 97 48.8 6 6 12 2 6 4
14 Nav 1 M 83 26.0 4 2 2 4 1 3
15 Nav 1 M 71 37.3 1 1 0 7 7 0
16 Nav 2 F 68 40.1 1 1 2 8 11 3
17 Nav 2 F 94 16.5 2 * * 3 * *
18 Nav 2 M 88 32.7 5 4 1 4 1 3
19 Nav 2 M 93 50.8 4 4 0 10 12 2
20 Nav 2 M 62 10.4 9 10 1 3 6 3

3
G Model
JINJ-4780; No. of Pages 8
4 O. Keast-Butler et al. / Injury, Int. J. Care Injured xxx (2011) xxx–xxx
Fig. 1. Computer navigation setup for ‘‘Nav 1’’ and ‘‘Nav 2’’ groups illustrating the
patient’s affected (right) and normal (left) lower limbs, active infrared trackers,
radiolucent operating table, C-arm system, image intensifier, and navigation system
at the foot of the operating table. The ‘‘Fluoro’’ group was assessed with a
conventional setup that was identical to that shown here, except without the use of
the computer navigation unit and trackers. The healthy limb was moved out of the
field of view during image acquisition.
the lesser trochanter was defined on antero-posterior/anterolat-
eral images and used in conjunction with the femoral head centre
to define a plane relating to the femoral neck anteversion.
Outcome measures
Following completion of surgery, cadavers had high-resolution
CT scans taken using a GE VCT 64 Slice Helical Scanner (General
Electric, Fairfield, CT, USA) with a 0.625-mm slice thickness. CT
scans were taken of the right and left legs from the superior aspect
of the pelvis to the proximal tibia,7 which allowed precise
comparison of repaired right femurs with matching normal left
femurs. Leg length was measured from the superior aspect of the
femoral head proximally to the articular surface of the medial
Fig. 2. Drawing of computer navigation setup showing top view of the operating
theatre and the manner in which various instrumentation was interfaced.
Please cite this article in press as: Keast-Butler O, et al. Computer navigation in the reduction and fixation of femoral shaft fractures: A
randomized control study. Injury (2011), doi:10.1016/j.injury.2011.08.020
femoral condyle distally, ensuring that the same landmarks were
measured for each leg. Rotation was calculated for each right femur
by measuring the posterior condylar axis of the knee and the angle
of the femoral neck and then comparing this with values for the
normal left limb. Scans were examined by a musculoskeletal
consultant radiologist who was blinded to the mode of treatment.
Statistical analysis
Fluoro group specimens were repaired and assessed using CT
scans only. Nav 1 and Nav 2 specimens were repaired using
computer navigation, but were then ‘double-checked’ after surgery
using CT scans. Differences between groups with respect to length
and rotation, comparing continuous data, were compared with
analysis of variation (ANOVA) with p < 0.05. Absolute values
ignoring negative signs were used, such that 10 mm shortening or
10 mm lengthening was considered the same deformity to
facilitate statistical analysis. Differences between groups with
respect to length and rotation were converted to categorical data
with cases of rotational deformity >158 or length abnormality
>10 mm defined as malreduction. Effect of age and fracture defect
size on navigated reduction accuracy was assessed using Pearson’s
correlation coefficient. Relationships were analysed between
length and rotational errors to investigate if errors in length for
individual cases were related to the respective rotational error.
One-tailed post hoc power analysis was performed for absolute
values from CT measurements to determine if there were enough
specimens per group to detect all actual statistical differences
present, that is, to avoid type II error.
Results
Quantitative results for alignment
The control group (Fluoro, Case 1–5) and study groups (Nav 1,
Case 6–15; Nav 2, Case 16–20) were not statistically different with
respect to age (mean 81, 80.8 and 81 years, respectively)
(p = 0.999). They were also not statistically different with respect
to mean fracture defect sizes (26.0, 29.5 and 30.1 mm, respective-
ly) (p = 0.905).
For leg length discrepancy, comparing the Fluoro, Nav 1 and Nav
2 specimens using only CT scans showed a statistical difference
(p < 0.023) (Table 1, Fig. 4). Similarly, comparing the CT-based
Fluoro results vs. computer navigation measurements on the Nav 1
and Nav 2 specimens also showed a difference (p < 0.016) (Table 1,
Fig. 5). The most accuracy with respect to length was seen in the
Nav 1 group with a mean length discrepancy of 3.6 mm (CT) and
2.56 mm (computer navigation). The least accuracy was in the
Fluoro control group with a mean deformity of 9.8 mm.
For rotational discrepancy, there were no statistical differences
when Fluoro, Nav 1 and Nav 2 specimens were assessed using only
CT scans (p = 0.650) (Table 1, Fig. 6). Similarly, comparing the CT-
based Fluoro results vs. computer navigation measurements on the
Nav 1 and Nav 2 specimens also showed no difference (p = 0.920)
(Table 1, Fig. 7).
A correlation was identified between cadaver age and length
error after excluding one outlier (r = 0.703; p < 0.012). No
correlation was seen comparing age with rotational error or
fracture size with length or rotational discrepancy. No correlation
was seen between length errors and rotational errors.
Post hoc power calculations for CT-based measurements yielded
leg length discrepancy powers of 68% (Nav 1 vs. Fluoro) and 55% (Nav
2 vs. Fluoro), as well as rotational deformity powers of 6% (Nav 1 vs.
Fluoro) and 21% (Nav 2 vs. Fluoro). To reach the 80% power
recommended for a good study design, leg length discrepancy
comparisons would have required only eight specimens per group
G Model
JINJ-4780; No. of Pages 8
O. Keast-Butler et al. / Injury, Int. J. Care Injured xxx (2011) xxx–xxx
Fig. 3. Computer navigation screens showing typical malunion at the femur’s diaphyseal fracture site. Anteroposterior and lateral images are shown.
Fig. 4. Length discrepancy compared to the normal leg using CT scans for Fluoro,
Nav 1, and Nav 2 specimens. There were statistical differences (p < 0.023). Error
bars are one standard deviation 95% confidence intervals.
Fig. 5. Length discrepancy compared to the normal leg using CT scans for Fluoro
specimens and computer navigation for Nav 1 and Nav 2 specimens. There were
statistical differences (p < 0.016). Error bars are one standard deviation 95%
confidence intervals.
Please cite this article in press as: Keast-Butler O, et al. Computer navigation in the reduction and fixation of femoral shaft fractures: A
randomized control study. Injury (2011), doi:10.1016/j.injury.2011.08.020
5
(for Nav 1 vs. Fluoro) and 10 specimens per group (for Nav 2 vs.
Fluoro). However, rotational deformity comparisons would have
required 9000 specimens per group (for Nav 1 vs. Fluoro) and 50
specimens per group (for Nav 2 vs. Fluoro), which would not have
been feasible logistically or financially.
Qualitative results for technique
Obtaining perfect antero-posterior and lateral fluoroscopic
images of the broken and normal femurs was difficult in this
investigation. Multiple attempts were required to get comparable
images. It was not clear how small errors in rotation of images
would affect the overall accuracy of the navigation software. This
was beyond the scope of this study, but it is important to know
how perfect the fluoroscopic images need to be to minimise time
and radiation exposure spent obtaining them. We did not compare
total fluoroscopic imaging time or operating time in treatment and
control groups. This may be significant in live surgery, particularly
if patients are unstable and require expeditious treatment.
Trackers were attached to the patient using invasive methods. A
scatter plot comparing length with rotation demonstrated no
correlation (R2 = 0.08). If our trackers moved, errors would be
expected to occur with respect to both length and rotation. There
was no correlation between the errors, suggesting that tracker
movement does not appear to be the cause of errors seen in
fracture reduction in the navigation group.
Discussion
Our study showed statistically improved accuracy in length
reduction in the navigated femoral nails in comparison with the
conventional femoral nails. Rotational discrepancy was similar in
all groups with no significant differences. In the 15 cases treated
using computer navigation, one case still had a rotational
discrepancy >158 compared with CT measurements.
We reviewed prior studies that assessed postoperative rotation
and limb length malunion following intramedullary nailing,
G Model
JINJ-4780; No. of Pages 8
6 O. Keast-Butler et al. / Injury, Int. J. Care Injured xxx (2011) xxx–xxx
Fig. 6. Rotational discrepancy compared to the normal leg using CT scans for Fluoro,
Nav 1, and Nav 2 specimens. There were no statistical differences (p = 0.650). Error
bars are one standard deviation 95% confidence intervals.
measured with ultrasound or CT (Table 2). The leg length and
rotational discrepancies from our combined Nav 1 and Nav 2 data
using our post-surgery CT ‘double-check’ measurements on these
specimens (Table 1, Columns 6 and 9) were substantially smaller
than a sizable portion of prior results.
In a prospective randomised study of 87 patients, Stephen et al.5
compared fracture-table and manual traction with relation to
reduction quality, operative time, complications and functional
status. They obtained rotational deformity >108 at rates of 34%
(fracture-table) and 20% (manual traction). They also reported leg
length discrepancy >10 mm at rates of 2% (fracture-table) and 7%
(manual traction). In comparison to these data, our navigation
method had a 27% rotational deformity rate, but a 0% leg length
discrepancy rate.
In another prospective randomised trial, Tornetta and Tiburzi11
measured axial and rotational alignment in patients with
antegrade (n = 38) and retrograde (n = 31) nailing. Patients with
unstable fracture patterns (n = 33) had postoperative CT scans to
measure the accuracy of reduction in relation to length and
rotation. A retrograde technique resulted in less accurate reduction
in terms of length and rotation. They reported a 24% rate for
rotational deformity >108, which was slightly superior to our 27%
rate for navigation.
In their prospective randomised study, Deshmukh et al.12
compared conventional reduction techniques, with techniques
based on aligning the fragments by matching fluoroscopic images
of the normal and fractured leg intra-operatively. Ten patients with
unstable fractures were randomised to the study groups (n = 5 in
each group). Postoperative CT scans were performed to measure
rotational reduction. Accuracy of reduction in the fluoroscopically
reduced group was significantly better than in controls (mean
rotational error 4.18 vs. 12.468, p = 0.016). They reported a 20% rate
for rotational deformities >158, which was higher than our 13% rate.
Jaarsma et al.13 investigated rotational malalignment following
intramedullary femoral nailing in 76 of 112 patients treated for a
femoral fracture with a locked intramedullary nail, using a traction
Table 2
Summary of prior studies investigating length and rotational malunion following intramedullary nailing with either ultrasound or CT examination.
Reference Type of study design Number of cases
Stephen et al.5 Randomized control trial 42 fracture table
Stephen et al.5 Randomized control trial 45 manual traction
Tornetta and Tiburzi11 Randomized control trial 33 unstable fractures
Deshmukh et al.12 Randomized control trial 10 fractures
Jaarsma et al.13 Observational cohort 76 fractures
Braten et al.14 Observational cohort 110 fractures
Salem et al.15 Observational cohort 41 fractures
Please cite this article in press as: Keast-Butler O, et al. Computer navigation in the reduction and fixation of femoral shaft fractures: A
randomized control study. Injury (2011), doi:10.1016/j.injury.2011.08.020
Fig. 7. Rotational discrepancy compared to the normal leg using CT scans for Fluoro
specimens and computer navigation for Nav 1 and Nav 2 specimens. There were no
statistical differences (p = 0.920). Error bars are one standard deviation 95%
confidence intervals.
table. Patients underwent clinical examination to measure detect-
able deformities and CT scans to measure rotational and leg length
deformities. The article concluded that 28% of patients had
significant rotational deformity >158 following intramedullary
nailing, causing difficulties with challenging activities. It also
demonstrated the inaccuracy of clinical examination in determining
rotational deformity. In comparison to these results, we achieved a
13% rate for rotational deformity >158 using computer navigation.
A retrospective review by Braten et al.14 of 112 consecutive
femoral shaft fractures treated with locked femoral nails was
performed, and 110 cases were clinically reviewed. Twenty-six
fractures were unstable (Winquist grade 3–4). Clinical examina-
tion and ultrasound techniques were used to measure length and
rotational deformity. Twenty-one cases (19%) had torsional
deformity >158, and 26 cases had torsional deformity from 108
to 148. They concluded that only rotational deformities >158 would
cause problems. In comparison, we obtained a 13% rate for
rotational deformity >158 with computer navigation. Moreover,
although Braten and colleagues report only a 9% rate for leg length
discrepancies >10 mm, this was still much higher than our 0% leg
length error rate for discrepancies >10 mm using navigation.
A retrospective cohort study by Salem et al.15 reviewed patients
with a distal third femoral shaft fracture treated with antegrade or
retrograde nailing. Fifty fractures met the inclusion criteria and 41
patients were reviewed, with 20 in the antegrade nailing group and
21 in the retrograde nailing group. Three (15%) antegrade nails and
five (23%) retrograde nails had rotational deformity >158, with a
mean rotational deformity of 9.858 and 98, respectively. However,
we generated a 13% rate for rotational deformity >158 using
navigation. Their leg length difference was a mean of 10.45 and
12.28 mm in the antegrade and retrograde groups, respectively,
with an overall 27% leg length error rate for discrepancies >15 mm.
By contrast, our leg length error rate using navigation was 0% for
discrepancies >15 mm.
Weil and co-workers assessed the accuracy of a computer
navigation system to assist in multiplanar fracture reduction in
Method of assessment Rotational Length discrepancy
of length/rotation deformity
CT 34% > 108 2% > 10 mm leg length difference
CT 20% > 108 7% > 10 mm leg length difference
CT 24% > 108 n/a
CT 20% > 158 n/a
Ultrasound 28% > 158 n/a
CT 19% > 158 9% > 10 mm shortening
Ultrasound 20% > 158 27% > 15 mm leg length difference
G Model
JINJ-4780; No. of Pages 8
O. Keast-Butler et al. / Injury, Int. J. Care Injured xxx (2011) xxx–xxx
open and blind closed procedures.16 They used cadaveric femurs
fixed in a simulator and optically tracked. Even for blind, closed
reductions, the accuracy of restoration of femoral length was
1.2  0.4 mm for a simple fracture and 1.9  1.8 mm for a segmental
fracture. Rotational accuracy was 1.7  1.98 and 2.5  1.88, respec-
tively. Open reduction yielded similar results. Their apparently
superior findings compared with our results and those of some other
studies may be due to their use of isolated femurs, whereas we used
whole body cadavers with intact surrounding tissues and structures
as would be done clinically, which would complicate body position-
ing on the radiolucent operating table, tracker placement, nail
insertion and so on.
Our study had several drawbacks:
1. The fracture model was chosen to demonstrate differences
between study and control groups. The proportion of unstable
fracture patterns in an epidemiological study has previously
been reported as 13%.2 The study population of comminuted
femoral shaft fractures, hence, is not representative of the
general population of femoral shaft fractures and results can
only be applied to unstable fracture patterns.
2. The cadavers frequently appeared to have severe osteopenia,
assumed to be related to their age, which was evident whilst
obtaining fluoroscopic images and also on reviewing CT scans.
Although this was not measured, it made landmark recognition
very difficult and may have diminished the accuracy of the
navigation technique.
3. Trackers were fixed to the uninjured leg with invasive pins to
ensure no errors occurred in registration of the normal leg. This
may not be acceptable clinically, but was used to remove a
potential source of error and to maximise accuracy. All reference
images of the normal femur could be obtained without moving
the limb. Trackers can either be attached to the patient using
invasive or non-invasive methods, or fixed to the operating table
directly, each with its pros and cons. Invasive fixation with
single and double pin configurations has previously demon-
strated movements with all tested configurations with an
applied load of 65 N.8 A non-invasive method of fixation of a
tracker to the thigh using a strap has been shown to have
significant movement.9 In the fixation of femoral neck fractures
using a traction table, fixation of the tracker to the operating
table has been demonstrated to be as accurate as invasive
fixation of the tracker to the patient,10 although this may be
unreliable if performing femoral nailing by freehand on a
radiolucent table. Given the nascent nature of our computer
navigation method, the effect of tracker motion still needs to be
investigated thoroughly.
4. The technique is not fully representative of the clinical scenario,
nor is it ready for clinical application in its present form. It is
cumbersome due to marker placement and pin insertion. We did
not consider swelling, pathological tissue tension, operative
time, blood loss or complications, which should be assessed if
the technique is to be useful as an adjunct method. Moreover,
although not strictly monitored, set-up time was longer using
the computer navigation system than the traditional fluoro-
scopic method. We accept this as a limitation, given that the
navigation method is still under development.
5. The study was slightly statistically underpowered for leg length
discrepancy comparisons, although statistical differences be-
tween navigation and controls were still evident. Moreover, the
study was greatly underpowered for rotational deformity
comparisons, but the large number of specimens required for
80% power was logistically and financially impractical.
6. It could be argued that inequalities in length around 10 mm are
not clinically important; however, our experience has shown
that patients begin to notice their own limb inequality starting
Please cite this article in press as: Keast-Butler O, et al. Computer navigation in the reduction and fixation of femoral shaft fractures: A
randomized control study. Injury (2011), doi:10.1016/j.injury.2011.08.020
7
at this point. For a patient group, an average 10-mm discrepancy
means that some patients in that group had a much larger
discrepancy, which would be clinically important. Thus, 10 mm
was a reasonable criterion for differentiating between accept-
able and unacceptable limb inequality in our study.
7. Although we performed free nailing with manual traction on a
fluoro table, the ‘gold standard’ is considered by some to be
nailing on a fracture table, which has provided excellent results
in the past.17,18 Yet, fracture tables can limit access to a patient
to treat any other injuries, require more time to transfer multi-
injured patients from table to table and have been associated
with well-leg compartment syndromes, pudendal nerve palsies
and perineum skin sloughs.17,19 Moreover, manual traction vs.
fracture-table traction for nailing of isolated diaphyseal
fractures can decrease operative time (119 vs. 139 min) and
minimise the number of cases of internal malrotation greater
than 108 (7% vs. 29%).5
8. In the non-locking nail era, many femoral rotational errors could
be successfully corrected by nailing using manual reduction.3
However, our study’s goal was not to compare the effect of
locked vs. non-locked nails or manual traction vs. fracture table
traction. Rather, given the use of locking nails with manual
traction in our investigation, we desired to compare two
visualisation methods, that is, traditional fluoroscopy vs.
computer navigation. As this latter method is developed further
in the future, the effects of these other factors should be
assessed.
Finally, there was still a large 27% rate of malrotations >108 as
seen from the postoperative CT measurements of the Nav 1 and
Nav 2 specimens, with two specimens showing large malrotation
>158 in the Nav 1 group, demonstrating how difficult an optimal
reduction is to achieve, even using an invasive technique in a
normal leg.
Conclusions
Our computer navigation techniques demonstrated superior
accuracy in the reduction of femoral shaft fractures compared with
a conventional fluoroscopic technique with respect to length
correction. Rotational error was similar in all groups. Our results
from computer navigation were superior overall compared with
standard methods in the literature.
Conflict of interest
Stryker Canada provided financial support for the lead author’s
fellowship studies when this study was performed. None of the
other authors obtained personal financial gain as a result of this
study. No other commercial funding was received.
Acknowledgements
The authors wish to acknowledge the assistance of John
Gilmore, Nathan Hale and Robert Scherer of Stryker Canada, Garret
Kort of Stryker USA and Max Singh and Helmut Rath of Stryker
Germany. We also thank Stryker Canada for financial support.
References
1. Arneson TJ, Melton 3rd LJ, Lewallen DG, O’Fallon WM. Epidemiology of diaphy-
seal and distal femoral fractures in Rochester, Minnesota, 1965–1984. Clin
Orthop Relat Res 1988;234:188–94.
2. Salminen ST, Pihlajamaki HK, Avikainen VJ, Bostman OM. Population based
epidemiologic and morphologic study of femoral shaft fractures. Clin Orthop
Relat Res 2000;372:241–9.
G Model
JINJ-4780; No. of Pages 8
8 O. Keast-Butler et al. / Injury, Int. J. Care Injured xxx (2011) xxx–xxx
3. Winquist RA, Hansen Jr ST, Clawson DK. Closed intramedullary nailing of
femoral fractures. A report of five hundred and twenty cases. J Bone Joint Surg
Am 1984;66:529–39.
4. Kempf I, Grosse A, Beck G. Closed locked intramedullary nailing. Its application
to comminuted fractures of the femur. J Bone Joint Surg Am 1985;67:709–20.
5. Stephen DJ, Kreder HJ, Schemitsch EH, Conlan LB, Wild L, McKee MD. Femoral
intramedullary nailing: comparison of fracture-table and manual traction. A
prospective, randomized study. J Bone Joint Surg Am 2002;84:1514–21.
6. Krettek C, Miclau T, Gruin O, Schandelmaier P, Tscherne H. Intraoperative
control of axes, rotation and length in femoral and tibial fractures. Technical
note. Injury 1998;29(Suppl 3):C29–39.
7. Tornetta 3rd P, Ritz G, Kantor A. Femoral torsion after interlocked nailing of
unstable femoral fractures. J Trauma 1995;38:213–9.
8. Mihalko WM, Duquin T, Axelrod JR, Bayers-Thering M, Krackow KA. Effect of
one- and two-pin reference anchoring systems on marker stability during total
knee arthroplasty computer navigation. Comput Aided Surg 2006;11:93–8.
9. Kendoff D, Bogojevic A, Citak M, Maier C, Maier G, Krettek C, et al. Experimental
validation of noninvasive referencing in navigated procedures on long bones. J
Orthop Res 2007;25:201–7.
10. Ilsar I, Weil YA, Joskowicz L, Mosheiff R, Liebergall M. Fracture-table-mounted
versus bone-mounted dynamic reference frame tracking accuracy using com-
puter-assisted orthopaedic surgery—a comparative study. Comput Aided Surg
2007;12:125–30.
Please cite this article in press as: Keast-Butler O, et al. Computer navigation in the reduction and fixation of femoral shaft fractures: A
randomized control study. Injury (2011), doi:10.1016/j.injury.2011.08.020
11. Tornetta 3rd P, Tiburzi D. Antegrade or retrograde reamed femoral nailing. A
prospective, randomised trial. J Bone Joint Surg Br 2000;82:652–4.
12. Deshmukh RG, Lou KK, Neo CB, Yew KS, Rozman I, George J. A technique to
obtain correct rotational alignment during closed locked intramedullary nailing
of the femur. Injury 1998;29:207–10.
13. Jaarsma RL, Pakvis DF, Verdonschot N, Biert J, Van Kampen A. Rotational
malalignment after intramedullary nailing of femoral fractures. J Orthop Trauma
2004;18:403–9.
14. Braten M, Terjesen T, Rossvoll I. Torsional deformity after intramedullary
nailing of femoral shaft fractures. Measurement of anteversion angles in 110
patients. J Bone Joint Surg Br 1993;75:799–803.
15. Salem KH, Maier D, Keppler P, Kinzl L, Gebhard F. Limb malalignment and
functional outcome after antegrade versus retrograde intramedullary nailing in
distal femoral fractures. J Trauma 2006;61:375–81.
16. Weil Y, Gardner MJ, Helfet DL, Pearle AD. Computer navigation allows for
accurate reduction of femoral fractures. Clin Orthop Relat Res 2007;460:185–91.
17. Browner BD. The science and practice of intramedullary nailing. 2nd ed. Baltimore:
Williams and Wilkins; 1996.
18. Küntscher G. In: Rinne HH, translator. Practice of intramedullary nailing.
Springfield, IL: CC Thomas; 1967.
19. Brumback RJ, Ellison TS, Molligan H, Molligan DJ, Mahaffey S, Schmidhauser C.
Pudendal nerve palsy complicating intramedullary nailing of the femur. J Bone
Joint Surg Am 1992;74:1450–5.