Twenty-four fresh-frozen human humeri were separated into four experimental
groups (six humeri in each group). The groups were selected for either bending
or torsion testing and fixation with either a proximal humeral intramedullary
nail or a locking proximal humeral plate. The donors included twenty-one men
and three women. The mean age of the donors at the time of death was
fifty-nine years (range, forty-two to seventy-four years) for the specimens
instrumented with an intramedullary nail and sixty years (range, forty-nine to
seventy-two years) for the specimens that received a plate. The bone quality
of the humeri was assessed with use of dual x-ray absorptiometry prior to
mechanical intervention. Bone mineral density was used to group the specimens
so that there was no significant difference in bone quality among the groups
(p = 0.902) (see Appendix). Anthropometric measurements were made for each
specimen; the total length of the specimen (head to distal condyles), head
circumference, and diaphyseal shaft circumference were recorded. No
significant difference in anthropometric measurements was detected among the
four experimental groups (p = 0.220) (see Appendix).
The distal part of the humerus was removed at the metaphyseal-diaphyseal
junction. Simulated fractures of the surgical neck were created at the base of
the greater tuberosity at a 10° oblique angle, directed in a medial and
inferior direction, with use of a thin reciprocating saw. A 10-mm wedge of
bone was excised to simulate comminution. The humeri were repaired with use of
either a titanium proximal humeral nail (PHN; Synthes, Paoli, Pennsylvania) or
a 3.5-mm locking compression plate for the proximal part of the humerus
(LCP-PH; Synthes).
For testing, the articular surface of the humeral head was fixed to the
mid-level of the anatomic neck in an aluminum tube with use of spiked screws
(screws with sharpened points) to hold the segment
(Fig. 1). For the bending
tests, each humeral shaft was cyclically loaded in a cantilevered fashion to
produce a varus bending moment of 0 to 7.5 Nm at the fracture site. A linear
actuator (Myostat Motion Control, Newmarket, Ontario, Canada) with a custom
LabVIEW control program (National Instruments, Austin, Texas) was used to load
the distal fragment with a u-shaped bone holder
(Fig. 1, A).
Displacement at the loading point was continuously measured by the encoder
built into the linear actuator. A force sensor (Honeywell International,
Columbus, Ohio) was used to monitor the contact force between the bone and the
u-shaped holder. For the torsion tests, each humeral shaft was cyclically
loaded to ±2 Nm of axial torque while angular rotation of the distal
fragment was continuously measured with use of the rotation encoder of the
servomotor (Danaher Motion, Wood Dale, Illinois)
(Fig. 1, B). A torque
sensor (S. Himmelstein, Hoffman Estates, Illinois) was used to measure the
corresponding torque.
For both testing methods, the specimens were continuously monitored to 5000
cycles. For the bending tests, mean displacement and maximum displacement were
calculated for each 1000-cycle increment and for the entire 5000-cycle testing
period. Mean displacement was defined as the mean magnitude displacement of
all cycles throughout each testing period. Maximum displacement was defined as
the largest magnitude displacement in each testing period. For the torsion
tests, mean angular rotation and maximum angular rotation were calculated for
each 1000-cycle increment and for the entire 5000-cycle testing period. Mean
angular rotation was defined as the mean magnitude rotation of all cycles
throughout each testing period. Maximum angular rotation was defined as the
largest magnitude rotation in each testing period. Mean displacement and mean
angular rotation data for each of the test groups was used to quantify
component loosening at each 1000-cycle increment. The cycle number, maximum
displacement, or maximum angular rotation at gross implant failure or cutout
was recorded for specimens that failed prior to the completion of 5000
cycles.
Stiffness of the bone-implant construct, or the slope of the linear portion
of the force-displacement curve, was calculated for each specimen.
Displacement versus force was plotted for each specimen tested in bending, and
angular rotation versus torque was plotted for each specimen tested in
torsion. The linear portion of each force-displacement or torque-angle curve
was determined with use of the least-squares method of linear regression. The
slope of the linear region of the force-displacement or torque-angle curve,
which is the stiffness of the bone-implant construct, was then determined for
each specimen.
After testing with cyclic bending, the specimens were loaded to the maximum
capacity of the testing apparatus (maximum, 125 mm of displacement), which did
not result in failure. The displacement and torque at maximum displacement
were recorded. Following cyclic torsion testing, the specimens were loaded to
failure. The torque, angle, and mode of failure were recorded.
Analysis of variance was used to test for significant differences between
the experimental groups in terms of bone mineral density and anthropometric
measurements. Analysis of variance with repeated measures was used to compare
the two methods of fixation at 1000-cycle increments over the 5000-cycle
testing period. The Student t test was used to compare the two methods of
fixation for independent measures. The Fisher exact test was used to evaluate
the absolute number of specimens that failed during cyclic loading for each
group. An alpha of 0.05 was used to test for significance.
In cyclic bending, the fractures fixed with the LCP-PH implant demonstrated
a significantly lower mean displacement over the entire 5000-cycle testing
period (p = 0.002) compared with fractures fixed with the PHN implant
(Table I). The mean
displacement increased significantly with increasing cycles for fractures
fixed with both the PHN and the LCP-PH implant (p < 0.001)
(Fig. 2). Maximum displacement
increased significantly with increasing cycles for fractures fixed with both
the PHN and the LCP-PH implant (p < 0.0001)
(Fig. 3). Over the entire
5000-cycle testing period, the fractures fixed with the PHN implant
demonstrated significantly more maximum displacement of the distal fragment
than fractures fixed with the LCP-PH implant (p = 0.001)
(Table I). No implant failure
or cutout was noted in either the PHN or LCP-PH group when subjected to cyclic
bending.
No significant difference was detected between the PHN group and the LCP-PH
group with regard to the maximum displacement of the distal fragment or the
load at maximum displacement. The mean stiffness (and standard deviation) in
cantilevered varus bending was 2.43 ± 1.16 N/mm for the PHN
bone-implant construct and 2.82 ± 0.95 N/mm for the LCP-PH bone-implant
construct; with the numbers tested, the two groups were not significantly
different.
In cyclic torsion, one of the specimens instrumented with the PHN was
discarded from testing secondary to a technical error with load to failure
being performed prior to cycling the specimen. For the 5000-cycle testing
period, the mean angular rotation was significantly greater in the PHN group
compared with the LCP-PH group (p = 0.040)
(Table II). In addition, the
maximum angular rotation was significantly greater in the PHN group compared
with the LCP-PH group (p = 0.013) (Table
II).
Three of the five specimens treated with the PHN implant failed in torsion
prior to completing the 5000-cycle testing protocol (p = 0.04) (see Appendix).
No implant failure or cut-out was noted in the humeri instrumented with the
LCP-PH implant when subjected to cyclic torsion. For the PHN-treated specimen
failures, with the numbers tested, there was no correlation between the bone
mineral density, head and shaft circumference, or proximal locking
screw-to-head diameter ratio and the cycle number at which failure occurred.
Only the mean angular rotation during the first 1000 cycles of torsion testing
in the PHN group was positively correlated with bone mineral density (r =
0.85, p = 0.032).
For the specimens that reached 5000 cycles with no implant failure or
cut-out, the PHN group failed at a significantly lower torque than the LCP-PH
group when loaded to failure in torsion (p = 0.009)
(Table III). In addition, the
LCP-PH implant created a significantly stiffer bone-implant construct than did
the PHN implant (p = 0.007) (Table
IV). Four of the five humeri treated with the PHN implant failed
by the proximal screw fracturing through, or cutting out of, the greater
tuberosity in the humeral head. One specimen instrumented with the PHN implant
failed in torsion with a spiral fracture that propagated through the distal
interlocking screw-holes. It is of note that the specimen that failed distally
had the smallest shaft circumference (72.78 mm) of all specimens tested on the
basis of anthropometric measurements. When loaded to failure in torsion, the
six specimens instrumented with the LCP-PH implant failed by the three distal
cortical screws pulling out of the distal fragment of the humerus. In every
specimen instrumented with the LCP-PH implant, the plate-screw construct
remained intact as did the fixation in the humeral head.
The optimal fixation technique for unstable fractures of the proximal part
of the humerus remains controversial. The results of this study suggest that a
locked proximal humeral plate is biomechanically more stable than a proximal
humeral intramedullary nail in a human cadaver model for the treatment of a
comminuted two-part fracture of the surgical neck.
Most unstable, comminuted proximal humeral fractures require surgical
management to restore useful shoulder function and are best treated with a
device to maintain length and alignment, such as an intramedullary device or a
plate-and-screw construct. Both techniques have advantages and disadvantages
clinically. The use of an intramedullary nail allows soft-tissue stripping to
be avoided, involves less operative time and blood loss, and provides the
ability to extend the instrumentation distally without increased exposure. The
potential disadvantages are shoulder pain and rotator cuff violation. A
plate-and-screw construct can provide an anatomic reconstruction with the
disadvantages of extensive exposure, soft-tissue stripping of bone fragments,
and increased intraoperative blood loss and time.
The bone mineral density values in this study are consistent with
previously published values of the density of the proximal part of the humerus
in both osteoporotic and nonosteoporotic cadaveric
specimens13. Paired
specimens were not used because of the potential influence of extremity
dominance on the bone mineral
density14,15.
A preliminary study that evaluated the LCP-PH plate compared with a 90°
angled blade-plate suggested that the LCP-PH was more stable in
torsion11. The
results of the current study suggest that the LCP-PH is a more stable
construct than the PHN in bending as well as in torsion. In the LCP-PH group,
there was significantly less distal fragment displacement in cantilevered
varus bending and significantly less angular rotation in torsion.
Of particular concern to us was the significant difference in failure rate
between the PHN and the LCP-PH implant when cycled in torsion, as 2 Nm is a
similar amount of torque to that experienced by the bone during activities of
daily living16.
Only the mean angular rotation during the first 1000 cycles of torsion in the
specimens that had the PHN was positively correlated with bone mineral
density. Interestingly, three of the four early catastrophic failures occurred
during the first 1000-cycle period. The rotational stability of the PHN used
in this study relied on the fixed-angle spiral blade in the humeral head and
the distal interlocking screws. However, it has been shown that the bone
quality, particularly at the greater tuberosity, is often poor relative to
other sites in the body, even in young active
patients17.
Therefore, it is not surprising to find that the rotational stability of such
a construct is influenced by bone mineral density. These results suggest that
the proximal humeral intramedullary nail may be a poor choice for use in
patients with osteoporosis.
In vivo load application on the proximal humeral fracture is complex,
comprising rotational, bending, and both compression and distraction forces
applied by the rotator cuff and surrounding musculature as well as by any
external loads applied to the
arm18. It is likely
that rotational and varus bending forces are the most likely to compromise
fracture
fixation16. Varus
bending represents a frequent physiologic displacement of the proximal humeral
fracture18. Siffri
et al. used bending moments of 0 to 7.5 Nm in a comparison of two
plate-fixation methods and found no significant differences or implant
failure11. We
hypothesized that the PHN implant would form a less stiff bone-implant
construct and would fail earlier. Therefore, varus bending of 0 to 7.5 Nm was
chosen as it was likely to elucidate differences between the LCP-PH and a less
stiff PHN bone-implant construct while remaining below load-to-failure levels.
Wheeler and Colville estimated that rotational moments from the infraspinatus
muscle during arm abduction were approximately 0.1 to 1.25 Nm and found that
cadaveric specimens failed during rotational loading at approximately 5.5 Nm
of torque16.
Therefore, to include the force-generating capacity of both the infraspinatus
and the subscapularis and to keep the rotational loading protocol at a
subfailure level, a rotational loading protocol of ±2 Nm was
chosen.
Lill et al. proposed that a sufficiently stable bone-implant construct was
one that was both flexible enough to unload the implant-bone interface and
rigid enough to minimize fracture
movements18. In the
current study, the stiffness of the bone-implant construct was examined as it
provides an accurate reflection of the biomechanical stability of the repair
in situ. The bone-implant constructs were found to have similar stiffness
results in cantilevered varus bending. In torsion, the PHN bone-implant
construct was found to be significantly less stiff than the LCP-PH. The rate
of early failure in the PHN implant in torsion could reflect the high moment
that is transmitted to the locking spiral blade-bone interface in this
implant.
The relatively simple loading pattern in both testing models, which is
unlikely to fully reproduce the complex nature of in vivo forces, is a
potential limitation of this study. In addition, in the torsion testing group,
as specimens in the PHN group were lost to failure, a beta error may have
occurred at the higher cycle numbers, causing an underrepresentation of the
level of significance between the two groups. Finally, following cyclic
testing, the specimens were loaded to the maximum capacity of the apparatus in
bending, which did not result in failure. It is possible that, had the
specimens been loaded to failure, significant differences may have been found
in maximum displacement of the distal fragment or in load at maximum
displacement.
There are several advantages and disadvantages to both the PHN and the
LCP-PH implants. Although the LCP-PH implant is biomechanically superior on
the basis of the results of this study, other clinical concerns, such as
operative time, may dissuade a surgeon from using this implant. In summary,
the LCP-PH demonstrated superior biomechanical characteristics compared with
the PHN when tested in both cyclic varus bending and torsion. The PHN
bone-implant construct was less stiff than the LCP-PH bone-implant construct.
The high rate of failure in torsion of the PHN-bone construct is concerning,
and further studies may be warranted to determine its clinical importance.
Details on the power analysis for this study as well as tables showing the
bone mineral density for each experimental group, the anthropometric
measurements between the experimental groups, and the mode and cycle at
failure in torsion for each specimen are available with the electronic
versions of this article, on our web site at
(go to
the article citation and click on "Supplementary Material") and on
our quarterly CD-ROM (call our subscription department, at 781-449-9780, to
order the CD-ROM). ?