Between September 1995 and December 1996, forty-four femora in
forty-two patients were treated at the University of Iowa Hospitals
and Clinics with a Food and Drug Administration-approved modular 316L
stainless-steel femoral nail (TriMax; Smith and Nephew, Memphis,
Tennessee) as part of a clinical evaluation of this device, which
had been released to a limited number of centers. The diagnosis
was an acute fracture in twenty-eight patients, a nonunion in eight,
and an impending pathologic fracture in eight.
We excluded the eight nails inserted to prevent impending pathologic
fracture because those patients had osteolysis secondary to tumor.
Another nail was excluded because it was associated with infection.
It was removed two weeks after insertion from a patient who had
been treated for a gunshot wound in the femur. Another modular femoral
nail, which had been placed through a retrograde approach, was excluded
because it broke at the proximal junction three months postoperatively.
It was exchanged for a second, larger modular femoral nail, and
this nail was included in the study. A third patient, who underwent
exchange nailing with a standard Russell-Taylor nail (Smith and
Nephew) at eleven months because of a nonunion, was also excluded.
Finally, a patient from whom a modular femoral nail was routinely
removed at another hospital was excluded.
This left thirty-two intramedullary modular femoral nails in
thirty-two patients. Following approval from our institutional review
board, each of these patients was sent a letter inviting them to
return for radiographs, a physical examination, assessment of functional
outcomes, evaluation of thigh pain with a visual analog scale, determination
of serum chromium levels, and nail removal if desired. All patients
signed an informed-consent form to participate in the study, and
all were counseled regarding the osteolytic reactions and their
possible clinical importance. The costs of the clinic visit and
nail removal were covered by the orthopaedic department, the hospital,
and the manufacturer. Five patients refused the invitation to return,
so twenty-seven patients (nineteen with an acute fracture and eight
with a nonunion) were evaluated at a mean of twenty-one months (range,
twelve to twenty-nine months) following the index operation. Fourteen procedures
were performed through an antegrade approach and thirteen, through
an intra-articular retrograde approach. There were fifteen men and twelve
women with a mean age of thirty-two years (range, nineteen to fifty-four
years). Ten right and seventeen left femora were involved. Twenty-three fractures
were closed, and four were open. With use of the Gustilo and Anderson
system for open fractures2, three
fractures were classified as grade II and one, as grade III. The
nail was statically locked in twenty-three of the twenty-seven fractures
and in all but one of the acute fractures. Twenty-six of the twenty-seven
fractures and nonunions healed. One patient (Case 23) had a nonunion
after the femoral nailing; this healed following exchange nailing
(Table I).
Serial radiographs were evaluated by two observers (D.M.J. and
J.L.M.), who began with the radiographs of the injuries. With use
of the Orthopaedic Trauma Association classification for femoral
fractures3, six of the nineteen
acute fractures were classified as OTA 32A; nine, as OTA 32B; and
four, as OTA 32C. The eight nonunions, which included six hypertrophic
and two atrophic nonunions, were treated at a mean of sixteen months
(range, three to thirty months) following the injury. Seven of the eight
patients had been treated previously with an intramedullary nail,
and one had been treated with a plate and screws.
The location of the fracture in relation to the modular junctions,
the tightest isthmal cortical fit (measured from the outer border
of the nail to the inner cortex), and the distance between the junctions
of the nail and the inner border of the cortex (measured to the
nearest millimeter) were determined on the initial postoperative
radiographs for all patients. The fracture was located at the level
of one of the junctions in twelve patients, and these junctions were
excluded from further analysis.
Three types of reactionsæosteolysis, periosteal reaction,
and cortical thickeningæwere observed radiographically
at the levels of the junctions. Osteolysis was defined as a focal,
hemielliptical area of rarefaction at the endosteal cortex. The extent
of osteolysis was quantified by measuring the distance, in millimeters,
from the endosteal cortex to the most medial or lateral extent of
the radiolucency along a line perpendicular to the long axis of
the femur (Fig. 2).
Cortical thickening was determined by subtracting the greatest width
of the cortex at the level of the junction on the immediate postoperative
radiograph from the same measurement made on the follow-up radiograph.
Periosteal reaction was recorded as either present or absent. The
time of appearance of the radiographic findings and their location
(at the large or small junction) were recorded.
At the time of the follow-up visit, each patient had an examination
of the lower extremities. Twenty-five patients completed the Musculoskeletal
Function Assessment instrument4,5 and
marked a visual analog scale for thigh pain, on which 0 indicated
no pain and 10 indicated severe, constant pain. Two patients had
the nail removed before the time of the evaluation with the assessment
form and the pain scale. Blood samples from twenty-five patients
were assayed for serum chromium concentration with use of graphite
furnace Zeeman atomic absorption spectrophotometry according to
previously described protocols6.
The detection limit for chromium in serum was 0.03 ng/mL
(parts per billion). Values below the detection limit were assigned
a concentration of one-half the detection limit by convention6. Serum samples from two of the twenty-seven patients
had been collected improperly and thus were discarded.
All patients were offered nail removal, and twelve nails were
removed. The retrieved nails were analyzed with scanning electron
microscopy (model 5900; JEOL, Peabody, Massachusetts) to evaluate the
nature of the degradation process at the taper interface.
Peri-implant soft tissues were obtained by curettage of the medullary
canal at the level of the modular junctions in eleven patients.
The tissues were processed for histologic analysis, and 5-mm-thick serial
sections were cut and stained with hematoxylin and eosin. The sections
were examined with light microscopy for cellular infiltrates and
the presence of particulate debris. The elemental composition of
corrosion products at the modular junctions of the nails and in
the histologic sections was determined with electron microprobe
analysis (JEOL model JXA 8900) with use of previously described
techniques7.
A control group of sixteen patients in whom an acute femoral
fracture had been treated with a standard one-piece 316L stainless-steel
Russell-Taylor intramedullary nail (Smith and Nephew) during the same
time-period was also identified. The mean age (twenty-nine years;
range, eighteen to fifty-two years) and duration of follow-up (twenty-two months;
range, thirteen to forty-two months) of these patients were similar
to those of the patients treated with the modular femoral nail.
These control patients were selected on the basis of the time of
implantation so that the duration of follow-up would be similar
to that of the patients with the modular nail. Ten nails had been
inserted in an antegrade fashion and six, in a retrograde fashion. Blood
samples were obtained for measurement of serum chromium concentration.
The pain analog scale and the assessment form were completed. Two
of the nails in the control group were removed. Curetted material
from the femoral canal was analyzed in the same fashion as it was
in the group treated with the modular femoral nail. We compared
the patients in the control group with the patients treated with
the modular nail with regard to serum chromium levels, scores on
the pain analog scale, and scores determined from the assessment form
and its subscales.
The serum chromium levels in a previously defined6 control group of twenty-one patients
without a femoral implant were compared with the levels in the patients
with the modular femoral nail and those in the control group with
the one-piece implant. The group consisted of twelve women and nine men
with a mean age of fifty-nine years (range, thirty-five to seventy-five
years).
The patients with the modular femoral nail were divided into
two groups on the basis of the severity of the radiographic findings.
For this purpose, mild or absent osteolysis was defined as <2
mm (twelve patients), and severe osteolysis was defined as 2 mm
(fifteen patients). Cortical thickening was considered to be mild
or absent if it was <5 mm (sixteen patients), and it was
considered to be severe if it was 5 mm (eleven patients). Periosteal
reaction was either present (ten patients) or absent (seventeen
patients). Eighteen patients with severe osteolysis, cortical thickening,
or a periosteal reaction at either junction were assigned to group
1, and nine patients with minimal or no reaction were assigned to
group 2. These groups were used to determine if there was an association
between the more severe radiographic reactions and the following
independent variables: age, gender, fracture type, location of the
fracture, location of the modular junction, nail size, method of
insertion, tightest isthmal fit of the nail overall and at the large
and small modular junctions, proximity of the modular junction to
the cortex, dynamic or static locking, and indication for the nailing
(an acute fracture or a nonunion). Serum chromium levels and levels
of thigh pain were also compared between the two groups.
Group 1 and group 2 were compared with use of the Fisher exact
test for the categorical variables (gender, method of insertion,
indication for the nailing, type of locking, and type and location
of the fracture). A two-sample t test was performed to compare the
groups with regard to the age at surgery, and the Wilcoxon two-sample
test was carried out to compare the groups with regard to pain,
nail size, tightest overall fit, and tightest fit at the large and small
modular junctions. The presence of severe findings (osteolysis only,
periosteal reaction only, cortical thickening only, or a combination
of the three) was compared between the two locations (subtrochanteric
and distal) within the same patient with use of the McNemar test.
Intergroup comparisons for significant differences in serum chromium levels
were performed with use of the Kruskall-Wallis test, followed by
a multiple comparison test based on rank sums if the Kruskall-Wallis
test indicated a significant difference at the p < 0.05
level.
The total scores and the subscale scores on the Musculoskeletal
Function Assessment were calculated according to the instructions
for administration. This outcomes questionnaire is based on a standardized
total score on a scale of 0 to 100, with 0 indicating no disability
and 100 indicating total disability. Intergroup comparisons of the
scores were made between the modular nail group and the control
group with the one-piece nail with a series of Student t tests with
significance set at p < 0.05.
Twenty-three of the twenty-seven patients (thirteen of the fourteen
patients who had had antegrade nailing and ten of the thirteen who
had had retrograde nailing) had at least one type of reaction at
one or both junctions (Table II). More than one type of reaction
was present at the involved junctions in thirteen patients. Osteolysis occurred
in nineteen patients and at twenty-three modular junctions (Figs. 3-A and 3-B). The osteolytic
lesions were as large as 5 mm, with a mean size of 2.5 mm. The mean
time until the osteolysis was first observed was thirteen months
(range, three to twenty-nine months). The extent of the osteolysis
decreased over time in only three patients. In eight patients, the
lesions were either stable (six patients) or they increased (two patients)
over time. In another eight patients, the osteolysis was first noticed
on the most recent radiographs (at a mean of twenty-two months; range,
seventeen to twenty-nine months).
A periosteal reaction occurred in ten patients and at thirteen
junctions. Seventeen patients (twenty-one junctions) had cortical
thickening (Figs. 4-A and 4-B). The mean time until radiographic
observation was nine months (range, two to twenty-nine months) for
the periosteal reactions and fourteen months (range, five to twenty-seven
months) for the cortical thickening. The mean amount of cortical thickening
was 6.1 mm (range, 0 to 15 mm). The periosteal reactions preceded
cortical thickening in all except two patients. In these two patients,
an obvious periosteal reaction that had not been present on radiographs
made only several months earlier was seen at the most recent follow-up
evaluation, at nineteen and twenty-nine months. The mean amount
of cortical thickening following a documented periosteal reaction
was 7.2 mm (range, 0 to 15 mm).
All three findings occurred more frequently in the subtrochanteric
region than in the distal region, and this difference was significant
when only the severe findings were considered (p = 0.02).
In the patients who had had antegrade nailing nineteen of the thirty-two
reactions involved the large junction, whereas in the patients who
had had retrograde nailing nineteen of the twenty-five positive
findings occurred at the small junction; both of these junctions
were in the subtrochanteric region. With the numbers available,
there was no association between the method of insertion and positive
radiographic findings. Age, gender, fracture type and location,
nail size, tightest overall isthmal fit, tightest fit of the modular
junction to the cortex, static or dynamic locking, and whether the
procedure was performed for an acute fracture or a nonunion were not
found to be significant factors (p > 0.05), with the numbers
available, in the development of severe reactions. Four modular
femoral nails were dynamically locked (three in a femur with a nonunion). Six
of the eight junctions in these nails were associated with a reaction.
The mean serum chromium level (and standard deviation) in samples
drawn from twenty-five of the twenty-seven patients with a modular
nail was 1.04 ± 0.57 ng/mL (range,
0.12 to 3.12 ng/mL). This level was significantly higher
than the mean level in the control group with the one-piece nail
(0.26 ± 0.40 ng/mL; range, 0.015
to 1.25 ng/mL [nondetectable in five patients])
(p < 0.05) and the mean level in the control group without
an implant (mean, 0.05 0.06 ng/mL; range, 0.015 to 0.25
ng/mL [nondetectable in twelve patients])
(p < 0.01). The mean in group 1 (patients with the modular
nail who had more severe radiographic reactions) was 1.27 ± 0.73 ng/mL, and the mean in group 2 (patients
with the modular nail who had mild or no radiographic reactions)
was 0.53 ± 0.41 ng/mL. This difference
did not reach significance with the numbers available (p < 0.90).
The mean serum chromium level in group 1 was significantly higher
than the mean level in the control group with the one-piece implant
(p < 0.01) and in the control group with no implant (p < 0.01).
Group 2 had a significantly higher mean serum chromium level when
compared with the control group with no implant (p < 0.01)
but not when compared with the control group with the one-piece
nail (p = 0.37). The difference between the two control
groups was not significant (p = 0.74).
The score for thigh pain on the visual analog scale averaged
2.2 (range, 0 to 6) in the modular nail group and 0.9 in the control
group with the one-piece nail. With the numbers available, this
difference was not significant (p = 0.07). However, the difference
between the patients in group 1 (mean, 2.6; range, 0 to 6) and the
control group was significant (p = 0.03). There was no
significant association between the presence of thigh pain and the serum
chromium level (p = 0.45). However, six patients with a
modular femoral nail had a score for thigh pain of 5 on the scale
of 0 to 10. Five of these six patients had osteolysis of 2 mm, and
four had a chromium level of >0.9 ng/mL.
The mean total score derived from the Musculoskeletal Function
Assessment was 22.6 ± 16.5 in the modular
nail group and 16.5 ± 14.0 in the control
group with the one-piece nail. These scores and the subscale scores
were not found to be significantly different with the numbers available
(p = 0.25).
All of the modular junctions of the retrieved nails were grossly
intact on manual testing. Corrosion products were visible at the
modular junctions of eleven of the twelve modular nails (Fig. 5). Evidence of
fretting corrosion was apparent on scanning electron microscopic
analysis (Fig. 6).
The fretting scars ran parallel with the long axis of the nail and
were approximately 100 to 200 mm in length.
Backscattered electron micrographs of corrosion debris found
in the taper interface, deep inside the crevice, showed that the
debris consisted of plates of corrosion product overlaying the alloy.
The corrosion plates at this location were found to contain a high
quantity of molybdenum relative to the alloy composition. At the
opening of the crevice, the corrosion products were different in
character and were of two types. The first type was a brownish flaky
deposit with a high iron and oxygen content and lesser amounts of
chromium. The second was a greenish plate-like deposit rich in chromium,
phosphorus, and oxygen.
Tissue specimens curetted from the medullary canal at the levels
of the modular junctions consisted of fragments of fibrous membrane
and granulation tissue demonstrating chronic inflammation. The cellular
infiltrate included abundant, particle-laden macrophages and foci
of plasmacytes accompanied by fewer lymphocytes (Fig. 7). There were
two types of particles within the macrophages: golden-brown, hemosiderin-like
granules, 1 to 3 mm in size, and pale-green, plate-like particles,
1 to 5 mm in size. Larger particles of the plate-like material,
as large as several hundred micrometers, were also present, associated
with multinucleated giant cells. Energy-dispersive x-ray analysis
indicated that the hemosiderin-like granules were rich in iron and
oxygen, with occasional traces of chromium. The greenish, plate-like
particles were rich in chromium, phosphorus, and oxygen and had
relatively low amounts of iron. Both of these corrosion products
in the tissues were similar in composition to the corrosion products
isolated from the modular junctions of the rods, surrounding the
opening of the crevice. No metal debris was found in the femoral
canal membrane of the two patients from whom a one-piece nail had
been removed.
Peri-implant osteolysis, which is most commonly associated with
total hip replacement, has not, to our knowledge, been previously
reported in association with femoral intramedullary nails. We suspect that
metal debris from corrosion at the nail junctions was responsible
for the local radiographic findings in our patients. The elevated
serum chromium levels and the visible corrosion products at the
modular junctions and within the tissue adjacent to eleven of the
twelve retrieved devices support this assertion. The corrosion that
was observed at the taper interfaces of the retrieved nails was
similar in character to mechanically assisted crevice corrosion
found in total hip femoral stems. Both mechanical fretting and evidence
of corrosion were found in our study. The corrosion debris found within
the taper, at both deep and superficial locations, appeared to be
rich in elements that are less soluble in aqueous electrolyte (for
example, molybdenum and chromium) compared with the alloy composition.
Again, this finding is similar to that seen at other tapers8. Fretting crevice corrosion of the
taper interfaces dramatically increases the rate of release of corrosion
products as well as the rate of reduction reactions (for example,
consumption of oxygen) elsewhere on the surface. The potential of
the implant also becomes more negative than its at-rest potential.
Each of these factors can affect the performance of the implant.
Also, the degradation of the taper interface may mechanically compromise the
nail, increasing the risk of device fracture9.
Two types of corrosion products were observed in the peri-implant
tissues. These were similar in composition to corrosion deposits
at the opening of the taper crevice and similar to particles reported
in association with stainless-steel internal fixation devices10,11. The first of these products
consisted of iron and oxygen-rich particles. While these particles
are somewhat similar to hemosiderin deposits, as might be expected
at the site of internal fixation of a fracture, the enormous volume
of these particles, the time since the fracture fixation, and the
fact that traces of chromium were detected in many of these particles
strongly suggest that they were a result of stainless-steel corrosion.
The second of these products was quite similar in histologic and
gross appearance, as well as in elemental composition, to the chromium
orthophosphate hydrate-rich corrosion product reported in association
with both similar-metal and mixed-metal modular femoral total hip-replacement
components7,12-14.
There is evidence that these chromium orthophosphate hydrate-rich
corrosion products may cause osteolysis not only by increasing polyethylene
wear through a three-body mechanism but also through a cellular
response from macrophages leading to bone resorption through a separate
mechanism12,15,16. Metal debris
has been shown to stimulate osteoclast production leading to bone
resorption17. In the published
literature on arthroplasty-associated osteolysis, it has not been
possible to isolate the effects of particulate articular wear debris, cement
debris, and metallic corrosion debris. The modular femoral nail
provides a unique system whereby the local tissue effects of metallic
corrosion debris alone can be observed. Femoral osteolysis was seen
in association with metallic corrosion products in the absence of
polyethylene, polymethylmethacrylate, or other debris typically
seen with joint arthroplasty reconstructions. This observation provides
additional evidence that at least some cases of osteolysis following
total hip arthroplasty may be the result of the generation of metallic
corrosion products at modular junctions.
While corrosion is unusual with one-piece internal fixation devices13,18, a positive association has been
demonstrated between the presence of corrosion and tissue reaction
around multiple-piece stainless-steel implants19-21.
These findings have been noted at a plate-and-screw interface that
has a substantially different geometry than the tight-fitting taper
junctions of the modular femoral nail. Also, the junctions between plates
and screws are small, extracortical, and away from the bone. Therefore,
it is unlikely that the small amount of corrosion products that
may develop would stimulate a localized cellular reaction substantial
enough to cause a reaction of the same magnitude that we have seen.
The device reported on in the present study has two large, intracortical
junctions where such fretting and corrosion could have occurred
and stimulated the type of cellular response necessary to cause
osteolysis.
The corrosion process at modular junctions of femoral total hip
components has been hypothesized to occur as a result of a combination
of cyclic relative motion, known as fretting, and crevice corrosion, which
is the electrochemical process that occurs after the breakdown of
the passivating oxide film in an environment with a restricted flow
of oxygen13,20,22. Spindle-shaped
cortical thickening and scalloping osteolysis secondary to corrosion
and abrasion of cemented titanium femoral stems have recently been
reported at postoperative intervals (cortical thickening at twenty-four
months and osteolysis at thirty-five months) similar to those in
our study16. It is likely that
a similar mechanism of accelerated corrosion occurred at the taper
junctions in the three-piece nails in our patients. The cellular response
to the metal debris and corrosion products, in turn, resulted in
focal osteolysis. Except for the series of cemented titanium stems
discussed above16, it appears
that the osteolysis observed in association with modular femoral
nails occurred earlier than is typically reported for hip arthroplasty.
The reasons for this are not clear but could be related to a higher
local concentration of corrosion products, the unique chemistry
of the corrosion products, the proximity (and intraosseous location)
of the modular femoral nail junctions to the cortical surface of the
femur (modular head-neck junctions of femoral total hip components
are in an extraosseous location), or a combination of these factors.
In every case but two, periosteal reaction preceded cortical
thickening. It is likely, therefore, that periosteal reaction and
cortical thickening occur along a continuum. The presence of a periosteal
reaction was also strongly associated with osteolysis. The periosteal
reaction and cortical thickening possibly resulted from increased
bone formation through injury and repair following the cellular
mediated bone resorption that occurs as a result of the corrosion
products.
The association of these radiographic findings with thigh pain
is of clinical interest. The patients in group 1, who had the more
severe reactions, had a significantly higher score for thigh pain
than did our control patients with a one-piece nail (p = 0.03).
Thigh pain was the presenting symptom in the first two patients
whom we observed to have radiographic reactions. Both had a large
area of osteolysis and a periosteal reaction at modular junctions
well away from the healed fracture. Also, 20% of the patients
with the modular femoral nail described pain that limited activities
of daily living. In their series of patients in whom cortical thickening
and osteolysis developed following primary implantation of a cemented
titanium total hip replacement, Willert et al. described the early
onset of dull, constant thigh pain, which was also thought to be
associated with similar radiographic findings16.
In two of our patients with marked osteolysis and thigh pain
(Cases 7 and 19), nail removal resulted in relief of the thigh pain,
an observation that further implicates corrosion products as the
cause of symptomatic osteolysis.
The time of appearance of these findings has been unpredictable,
ranging from two to twenty-nine months after injury. Of the many
factors evaluated, only the location of a nail junction in the subtrochanteric
region appeared to have an association with the prevalence of severe
reactions. The increased stresses in this area may have resulted
in more extensive fretting, facilitating the process of mechanically
assisted crevice corrosion22.
Circulating levels of metals, particularly chromium, have been
associated with corrosion at modular junctions in patients with
cobalt-chromium-alloy implants6,12.
In our study, the mean level of serum chromium in the patients with
the modular femoral nail was significantly higher than that in our
control group with a one-piece stainless-steel intramedullary nail
(p < 0.01) and that in our control group without an implant
(p < 0.01). While the clinical impact of elevated serum
levels of chromium and their association with osteolysis have not
been fully defined, serum chromium levels may well serve as a marker for
the degree of fretting corrosion at the modular junctions. This
has also been suggested to be the case for fretting corrosion at
the modular junctions of cobalt-chromium-alloy total hip replacements6,12.
In summary, we concluded that a biological reaction to fretting
corrosion products around the junctions of this modular femoral
nail caused the radiographic findings of osteolysis, periosteal
reaction, and cortical thickening; the elevated serum chromium levels;
and the thigh pain in our patients. As a result of these findings,
we have abandoned the use of this device, and it was withdrawn from the
other institutions participating in the clinical evaluation. Surgeons
and designers of fracture implants must be certain that the benefits
of modularity are not offset by the problems related to load-bearing
metal-metal junctions. To realize the benefits of modularity, careful
preclinical laboratory evaluation of modular devices with use of
relevant in vitro simulations must be performed23.