Abstract
Background: High-energy trauma to the lower
extremity presents challenges with regard to reconstruction and rehabilitation.
Failed efforts at limb salvage are associated with increased patient
mortality and high hospital costs. Lower-extremity injury-severity scoring
systems were developed to assist the surgical team with the initial
decision to amputate or salvage a limb. The purpose of the present
study was to prospectively evaluate the clinical utility of five lower-extremity
injury-severity scoring systems.
Methods: Five hundred and fifty-six high-energy
lower-extremity injuries were prospectively evaluated with use of
five injury-severity scoring systems for lower-extremity trauma
designed to assist in the decision-making process for the care of
patients with such injuries. Four hundred and seven limbs remained
in the salvage pathway six months after the injury. The sensitivity,
specificity, and area under the receiver operating characteristic
curve were calculated for the Mangled Extremity Severity Score (MESS);
the Limb Salvage Index (LSI); the Predictive Salvage Index (PSI);
the Nerve Injury, Ischemia, Soft-Tissue Injury, Skeletal Injury, Shock,
and Age of Patient Score (NISSSA); and the Hannover Fracture Scale-97
(HFS-97) for ischemic and nonischemic limbs. The scores were analyzed in
two ways: including and excluding limbs that required immediate
amputation.
Results: The analysis did not validate the clinical
utility of any of the lower-extremity injury-severity scores. The
high specificity of the scores in all of the patient subgroups did
confirm that low scores could be used to predict limb-salvage potential.
The converse, however, was not true. The low sensitivity of the
indices failed to support the validity of the scores as predictors
of amputation.
Conclusions: Lower-extremity injury-severity scores
at or above the amputation threshold should be cautiously used by
a surgeon who must decide the fate of a lower extremity with a high-energy
injury.
High-energy trauma to the lower extremity presents challenges
with regard to reconstruction and rehabilitation. Failed efforts
at limb salvage are associated with increased patient mortality
and high hospital costs. The decision to amputate or salvage a severely
injured lower extremity is difficult. Attempts to quantify the severity
of the trauma and to establish numerical guidelines for the decision
to amputate or salvage the limb have been proposed by several authors1-7. Published lower-extremity injury-severity
scoring systems include the Mangled Extremity Severity Score (MESS)2,4; the Predictive Salvage Index (PSI)3; the Limb Salvage Index (LSI)6; the Nerve Injury, Ischemia, Soft-Tissue
Injury, Skeletal Injury, Shock, and Age of Patient (NISSSA) Score5; and the Hannover Fracture Scale-97
(HFS-97)8. The developers of these
scoring systems attempted to validate them by demonstrating high
rates of specificity and sensitivity in predicting limb salvage.
However, independent testing of some of these scoring systems has
not duplicated the success reported by the developers of these systems5,9-11. To our knowledge, none of the
scoring systems have been evaluated prospectively with use of a large
cohort of patients treated at multiple facilities.
We report the results of an independent, prospective evaluation
of five injury-severity scoring systems for lower-extremity trauma
designed to assist in the decision to amputate or salvage a severely
injured limb. The sensitivity, specificity, and area under the receiver
operating characteristic curve were calculated and were used to
assess the ability of the indices to predict the need for amputation
within six months after the injury. The scores were assessed two
ways: by including and excluding the limbs that required immediate
amputation. The limb-salvage scores were evaluated for all limbs
in the study population as well as for subgroups, including ischemic
limbs; type-IIIA, IIIB, and IIIC tibial fractures according to the
system of Gustilo et al.12; severe
distal tibial fractures (open pilon fractures or type-IIIB ankle
fractures); hindfoot fractures; and isolated soft-tissue injuries
to the lower extremity.
Five lower-extremity injury-severity scoring systems were evaluated
with use of data collected as part of a larger, multicenter study
(the Lower Extremity Assessment Project [LEAP]) that was designed
to compare clinical and functional outcomes following amputation
with those following salvage of the lower extremity after severe
trauma. The LEAP project is described first, followed by a more
detailed description of the methods relevant to the present analysis.
LEAP Study
LEAP is a prospective longitudinal study of 601 patients, sixteen
to sixty-nine years old (mean, thirty-six years old), who were admitted
to one of eight level-I trauma centers between March 1, 1994, and
June 30, 1997, for the treatment of high-energy trauma to the lower
extremity. Participating centers include Carolinas Medical Center
(Charlotte, North Carolina), The R Adams Cowley Shock Trauma Center
(Baltimore, Maryland), Harborview Medical Center (Seattle, Washington),
Wake Forest University Baptist Medical Center (Winston-Salem, North
Carolina), Vanderbilt University Hospital (Nashville, Tennessee),
Tampa General Hospital (Tampa, Florida), University of Southwestern Texas
Medical Center (Dallas, Texas), and Cleveland MetroHealth Medical
Center (Cleveland, Ohio). The study was approved by the Institutional Review
Board at each of the centers, and all of the patients consented
to project participation, including follow-up evaluations. High-energy
injuries to the lower extremity were defined as injuries resulting
in a traumatic amputation (a functionally severed limb) below the
distal aspect of the femur or as injuries associated with some risk
of amputation, including (1) Gustilo type-IIIB and IIIC tibial fractures,
(2) selected Gustilo type-IIIA tibial fractures, (3) dysvascular
limbs (knee dislocations, closed tibial fractures, or penetrating
wounds with vascular injury), (4) major soft-tissue injuries to
the tibia (degloving or severe crush or avulsion injuries), and (5)
severe injuries to the distal aspect of the tibia or injuries to
the foot (type-III pilon fractures, type-IIIB ankle fractures, or
severe open injuries to the midfoot or hindfoot). A more detailed
definition of high-energy traumatic injuries to the lower extremity
is included in the Appendix.
Patients were excluded from participation in the LEAP study if
(1) they were less than sixteen years old or more than sixty-nine
years old, (2) they had a documented psychiatric disorder, (3) they
had an associated moderate-to-severe injury to the central nervous
system, (4) they had a third-degree burn to the injured limb that
measured more than one handbreadth, (5) they had a prior limb amputation
or were nonambulatory before the injury, (6) the primary treatment
was received prior to admission to a participating trauma center,
(7) they did not speak English or Spanish, or (8) they lived too
far outside of the catchment area of the treatment center to return
for follow-up evaluations.
In order to participate in the LEAP study, each trauma center
agreed to use acceptable practices for the care of high-energy injuries
to the lower extremity. These practices included emergency d衲idement
of wounds, stabilization of fractures, and antibiotic coverage.
Attending surgeons were required to be present for the initial evaluation
of all patients in the LEAP study and to direct all treatment decisions.
A patient-eligibility review team was established to independently
review all of the cases to ensure the patient's eligibility for
the study and to confirm the Gustilo classification12. All limbs were graded at the time
of the index surgical procedure. The limbs entered into the limb-salvage
pathway were graded again at the time of definitive soft-tissue-closure
or delayed amputation. The final grading was based on the soft-tissue closure
requirements of the limb. The limbs of patients with eligible injuries
that required soft-tissue coverage other than by delayed primary
closure or split-thickness skin grafts were classified as Gustilo
type IIIB. The limbs that had soft-tissue coverage by delayed closure
and/or split-thickness skin-grafting were classified as Gustilo
type IIIA. A Gustilo type-IIIC classification was assigned only to
limbs that required a vascular procedure to maintain viability.
Severe open tibial fractures with arterial injury (such as laceration
of the isolated posterior tibial artery) and acceptable distal flow were
not included as type-IIIC injuries. The admission photographs of
the injured limb as well as the radiographs and the description
of the soft-tissue injury in the medical records were used to confirm or
contest a patient's eligibility to take part in the study. The principal
investigator at the study site and the LEAP study eligibility-review
team had to reach an agreement on the inclusion or exclusion of all
patients tentatively enrolled in the study, and they were required
to resolve all differences in initial fracture classifications.
Inclusion Criteria for the Analysis of Limb-Salvage Scores
The five lower-extremity injury-severity scoring systems (MESS,
PSI, LSI, HFS-97, and NISSSA) were evaluated by examining their
ability to predict the need for an amputation. The scores were evaluated
in two ways: including and excluding limbs that required immediate
amputation. The analysis included immediate amputations to facilitate
the comparison of the present study with previous investigations,
all of which included immediate amputations when the validity of
a specific scoring system was assessed. Amputations were defined
as immediate if they were performed, as the index procedure, within
the first twenty-four hours after the injury.
When immediate amputations are included, the performance of a
score is not accurately tested. Correctly predicting the amputation
of a limb that would never be entered into a limb-salvage pathway
because of insurmountable injury and/or patient factors (for example,
when a limb remains ischemic and thus is beyond the reconstruction capability
of the surgeon and/or when the patient's condition prohibits a limb-salvage
and reconstruction effort) provides little useful information regarding
the clinical utility of a limb-salvage score. Therefore, the utility
of the limb-salvage scores was also analyzed by evaluating only
limbs that underwent a delayed amputation, with the immediate amputations
excluded. A delayed amputation was defined as any amputation that
was performed within six months after the injury but only after
some other treatment (fracture stabilization, revascularization,
or soft-tissue coverage of the sites of previous operations) had
been attempted. All delayed amputations were performed sometime after
the first twenty-four hours following the injury. A successfully
salvaged limb was defined as an extremity that remained in the limb-salvage
and reconstruction pathway six months after the injury. Six months
was selected as the end point for this analysis because patients
who had amputation after that time would be most likely to have
had major complications or intolerance to additional reconstruction
efforts, or both. The amputation decision at this point is likely
to be made independent of the initial limb-salvage index score.
Six hundred and one patients (633 lower extremities) who had
sustained a high-energy traumatic injury to the lower limb were
enrolled in the LEAP study. Fifty-five traumatic amputations were excluded
from the present analysis. Twenty-two additional limbs were lost
to follow-up. Five hundred and fifty-six limbs in 539 patients were
available for analysis (Group A). Sixty-three of these limbs were
not entered into a limb-salvage pathway and were immediately amputated.
Of the remaining 493 limbs in 477 patients (Group B), fifty-four
had dysvascular injury or ischemia, 199 had a fracture of the proximal
third or the diaphysis of the tibia, 201 had a fracture of the distal
aspect of the tibia or the foot, and thirty-nine had an isolated
soft-tissue injury (Fig. 1Fig. 1).
The mean age of the patients in the LEAP study was thirty-six
years (range, sixteen to sixty-nine years); 77% (463) of the 601
patients were male. Sixty-four percent (382) of the patients were injured
as a result of a motor-vehicle, motorcycle, or pedestrian-related
accident. The average Injury Severity Score (ISS)13 was
11 points; only 108 patients (18%) had an ISS of more than 16 points.
Forty patients (67%) who required immediate amputation had an ISS
of less than 17 points.
Data Collection
Data for this analysis were collected from three sources: (1)
a prospective clinical assessment of the injury and its treatment,
(2) medical records, and (3) follow-up clinical assessments of the
patient at three and six months after the injury. The attending orthopaedic
surgeon completed the initial assessment of the injured limb as
close to the time of admission as possible. We did not record the
timing of the initial data-book completion and cannot address potential
bias introduced to the study by data entry after a limb-ablation
procedure. The evaluation by the attending surgeon was designed to
characterize the severity of the injury at the time of initial presentation
and to score the injury according to the lower-extremity injury-severity scores
under study. To complete the initial evaluation, the orthopaedic
surgeon documented the severity of the tibial injury according to
the open-fracture classification system of Gustilo et al.12, the Orthopaedic Trauma Association
classification of long-bone fractures14,
and the AO/ASIF classification of soft-tissue injury15. The surgeon was then presented
with the components and the levels within each component of the MESS,
PSI, LSI, and HFS-97 and was asked to choose the appropriate level
of each component for these indices. The NISSSA was developed after
the initiation of the LEAP study. Therefore, the components of the
NISSSA were not explicitly included as part of the initial fracture-classification
protocol. The NISSSA is similar to the MESS and was designed to
improve the performance of the MESS by adding nerve injury to the
scale and incorporating more detailed information about the muscle
and soft-tissue injuries. We were able to assign scores to all of
the NISSSA components and to identify a NISSSA score for each patient
with use of data (the MESS, PSI, LSI, HFS-97, open-fracture classification
according to the system of Gustilo et al., AO classification of
soft-tissue injury, and Orthopaedic Trauma Association classification
of long-bone fractures) from the existing prospective database.
Calculation of Scores and Data Analysis
The five lower-extremity injury-severity scores vary in terms
of the factors considered relevant to limb salvage; they are summarized
in Table ITable
I. With the exception of the HFS-97, the component levels of each
injury-severity score were totaled according to the instructions
specified by its developers. The HFS was revised after the beginning
of the LEAP project. It was not possible to sum the component levels
of the 1997 revision of the HFS according to the instructions of
its developers because the wound-microbiology component of the HFS-97
was not recorded for 99% (353) of the patients enrolled in the LEAP
study with a Gustilo type-III tibial fracture. Initial microbiological
analysis of wounds is not part of a routine evaluation of an injury
in the United States. Rather than not evaluate the HFS-97 at all,
we consulted with the developers and decided to calculate a modified
HFS-97 on the basis of the sum of the remaining components of the
scale. The overall scores for each lower-extremity injury-severity
scoring system were not tabulated in the study data books, nor were
individual patient scores ever revealed to the surgeon.
To examine the discriminant validity of the five lower-extremity
injury-severity scores, the sensitivity and specificity for predicting
amputation were calculated. The sensitivity (the probability that limbs
requiring an amputation will have limb-salvage scores at or above
the index threshold) is defined as the number of limbs amputated
with scores at or above the threshold divided by the total number
of limbs amputated in the six-month follow-up period. Specificity
(the probability that salvaged limbs will have limb-salvage scores
below the threshold) is defined as the number of salvaged limbs
with scores below the threshold divided by the total number of salvaged
limbs for six months after the injury. The recommended amputation-threshold
scores published for the MESS, LSI, PSI, and NISSSA were used in
calculating sensitivity and specificity. Since we calculated a modified HFS-97,
we could not use the threshold recommended by the developers. Instead,
we examined the sensitivity and specificity distribution of the modified
HFS-97 at each score and chose a threshold that yielded a high specificity
(98%) with the least reduction in sensitivity. In choosing the threshold
value, we erred on the side of a high specificity, since it is believed
that the consequences of amputating a limb that can be salvaged
are much worse than those following a failed salvage. Limbs that
scored at or above the index threshold were predicted to be unsalvageable;
limbs that scored below the threshold were predicted to be salvageable.
The ability of the lower-extremity injury-severity scores to
discriminate between limbs that would be salvaged and those that
would not be salvaged (as defined by the need for an amputation)
was further evaluated by constructing receiver operating characteristic
curves and calculating the area under these curves16. A receiver operating characteristic
curve plots the sensitivity of the index by its false-positive fraction (1
- specificity) over the entire range of possible decision thresholds.
Each point on the curve represents the sensitivity/specificity pair
associated with a particular decision threshold. The area under
this curve provides a single number that summarizes the performance
of the index over the entire range of possible decision thresholds.
Thus, it provides a measure of index discrimination that is not
dependent on a specific decision threshold and therefore complements
the measures of sensitivity and specificity described above. Areas
under the receiver operating characteristic curve range from 0.50 (indicating
that the index performs no better than chance in discriminating
between groups) to 1.0 (indicating perfect discrimination). For
example, an area under the receiver operating characteristic curve
of 0.80 means that, in eighty of 100 cases, a randomly selected
individual from among those with delayed amputation will have a
larger value on the limb salvage index than a randomly chosen individual
among those whose limb is successfully salvaged. Generally, areas
under the curve of less than 0.70 represent poor discrimination;
values of 0.70 to 0.90, moderately good discrimination; and values
greater than 0.90, excellent discrimination17-19.
A chi-square test was used to determine whether the areas under
the receiver operating characteristic curve were significantly different
between two indices.
The sensitivity, specificity, and area under the receiver operating
characteristic curve were calculated for the MESS, LSI, and PSI
of the limbs in Group A (immediate amputations included) and of those
in Group B (immediate amputations excluded). The performance of
these indices was also examined for subgroups of ischemic and nonischemic
limbs. The HFS-97 and the NISSSA were excluded from this analysis,
as these indices were developed for evaluating limb viability associated
with open tibial fractures only. The analysis was repeated but was
restricted to limbs in the LEAP study that had an open tibial fracture.
This provided a comparison of all five indices, including the HFS-97
and the NISSSA. These comparisons were made for all tibial fractures
in the LEAP study and separately for Gustilo type-IIIB and type-IIIC fractures.
Comparison of the MESS, PSI, and LSI
The numbers of amputations (immediate and delayed) and extremities
that were salvaged successfully are presented in Table IITable II according
to the predicted outcome as determined with the threshold scores
of the MESS, PSI, and LSI. One hundred and forty-nine (26.8%) of
the limbs were amputated; sixty-three amputations were immediate
(11.3%), and eighty-six were delayed (15.5%). Four hundred and seven
limbs remained in the limb-salvage pathway for six months.
Of the eighty-six delayed amputations, sixty-eight were performed
during the initial hospitalization; twelve, during the first three
months after the injury; and six, during the second three months. Eighty-one
limbs with scores below the recommended limb-salvage threshold for
the MESS, PSI, and LSI underwent amputation. A number of limbs with
scores at or above the amputation threshold were salvaged: thirty-six
with scores at or above the MESS threshold, fifty-five with scores
at or above the PSI threshold, and twelve with scores at or above
the LSI threshold.
With use of the data in Table IITable II, the sensitivity, specificity,
and area under the receiver operating characteristic curve for the MESS,
PSI, and LSI were determined for Group A (including immediate amputations)
and Group B (excluding immediate amputations). The performance of
the limb-salvage indices was also determined for ischemic and nonischemic
limbs and for nonischemic limb-injury subgroups: Gustilo type-IIIA
and IIIB fractures of the middle or proximal aspect of the tibia,
Gustilo type-IIIA and IIIB fractures of the distal aspect of the
tibia and LEAP foot injuries, and LEAP soft-tissue injuries only.
The results are presented in Tables IIITables III and IVIV.
The MESS, PSI, and LSI demonstrated a high specificity (91%,
87%, and 97%, respectively) but a low sensitivity (46% each) for
the Group-A limbs (Table IIITable III). Analysis of the area
under the curve suggests that these indices have only moderately
good discrimination in their ability to predict salvage or amputation
of the limb. When only the eighty-eight ischemic limbs were considered,
the performance of the MESS deteriorated to a sensitivity of 72%,
a specificity of 62%, and an area under the curve of 0.66 (poor
discrimination). The performance of the MESS in the evaluation of
the Gustilo type-IIIA and IIIB proximal and mid-tibial fracture
subgroup was similar: a sensitivity of 22%, a specificity of 92%,
and an area under the curve of 0.68. The LSI performed significantly
better than did the MESS and the PSI for the entire cohort and for
the ischemic group (p < 0.05).
Table IVTable
IV presents the same analysis, excluding the sixty-three immediate
amputations. While the specificity for the entire cohort remained
unchanged, the sensitivity declined; the MESS had a sensitivity
of 27%; the PSI, of 36%; and the LSI, of 26%. When only the fifty-four
ischemic limbs in this subgroup were evaluated, the MESS demonstrated
a sensitivity of 55%, a specificity of 62%, and an area under the
curve of 0.59 (poor discrimination). Only the LSI performed with
moderately good discrimination in all groups and subgroups.
Comparison of All Five Severity Scores for
Open Tibial Fractures
The numbers of amputations (immediate and delayed) and successfully
salvaged limbs are presented in Table VTable V according to the predicted
outcome as determined with the thresholds of the MESS, PSI, LSI, NISSSA,
and HFS-97. One hundred (28%) of the 357 limbs with a Gustilo type-III
tibial fracture (eighty-four with a type-IIIA fracture [LEAP], 214 with
a type-IIIB fracture, and fifty-nine with a type-IIIC fracture)
were amputated (forty-five immediately and fifty-five after a delay).
Two hundred and fifty-seven limbs remained in the limb-salvage pathway
for six months. Amputated limbs were scored at or above the index
threshold on thirty-three NISSSA, thirty-seven HFS-97, forty-five MESS,
forty-seven PSI, and fifty-one LSI evaluations. Amputated limbs
were scored below the recommended threshold on sixty-seven NISSSA, sixty-three
HFS-97, fifty-five MESS, fifty-three PSI, and forty-nine LSI evaluations.
According to the data in Table VTable V, the sensitivity, specificity,
and area under the receiver operating characteristic curve for the MESS,
PSI, LSI, NISSSA, and HFS-97 were determined for the limbs in Group
A and Group B that had a Gustilo type-IIIA (LEAP), IIIB, or IIIC
tibial fracture and then were calculated separately for the type-IIIB
limbs and the type-IIIC limbs.
When the lower-extremity injury-severity scores were applied
to all 357 open tibial fractures in the LEAP study, the LSI performed
significantly better (a specificity of 97%, a sensitivity of 51%,
and an area under the curve of 0.85 ) than did the other scoring
systems (p < 0.05) (Table VITable VI). When only the fifty-nine
type-IIIC tibial fractures were considered, the LSI performed significantly better
than did the PSI, MESS, and NISSSA, while the HFS-97 performed better
than did the MESS and the NISSSA (p < 0.05). The MESS was only 69%
specific and 78% sensitive, with an area under the curve of 0.68
(poor discrimination).
When the immediate amputations were eliminated from the analysis
(Table VIITable
VII), the performance of the scores diminished. The LSI performed
significantly better than did the MESS, PSI, NISSSA, and HFS-97
(p < 0.05), with a 97% specificity, a 29% sensitivity, and an
area under the curve of 0.79 (moderately good discrimination). No
significant difference was found between the scores when only the
type-IIIC fractures were considered in the analysis. The MESS was
only 69% specific and 57% sensitive, with an area under the curve
of 0.62 (poor discrimination) for the type-IIIC limbs.
The lower-extremity injury-severity scores were developed to
assist the surgeon in making the initial decision of whether to
salvage or amputate an injured limb. Ideally, a trauma limb-salvage
index would be 100% sensitive (all amputated limbs with trauma limb-salvage
scores at or above the threshold) and 100% specific (all salvaged
limbs with scores below the threshold), and the receiver operating
characteristic curve would have an area of 1 (perfect accuracy).
Few clinical tests perform ideally. Depending on the reason for
the test (to screen or to aid in the decision to perform irreversible treatment),
trade-offs between acceptable levels of sensitivity and specificity
must be determined. In the decision to amputate, high specificity
is clearly important to ensure that only a small number of salvageable
limbs are incorrectly assigned a score above the decision threshold.
Sensitivity is also important, however, to guard against inappropriate delays
in amputation when the limb is ultimately not salvageable. High
rates of complications, including death, have been reported in these
latter cases20.
With the exception of the HFS-97, the development of the lower-extremity
injury-severity scores evaluated in the current study has been flawed
by retrospective design and small sample sizes. In addition, component
selection and weighting in all of the indices were affected by the
established clinical bias of the index developers. The MESS, NISSSA, and
HFS-97 all heavily weigh the results of initial neurological examination,
with the assumption that an acute sensory impairment correlates
with diminished limb-salvage capacity and that the initial examination
demonstrates the final deficit. However, ischemia, contusion, stretch,
or compression can cause transient neurological injury. With use
of the LSI, the neurological deficit is scored on the basis of anatomical
nerve findings.
Predictive Salvage Index (PSI)
The PSI was introduced by Howe et al.3 to
assess the condition of patients with combined orthopaedic and vascular
injuries. The intent of the PSI was to help prevent the attempted
salvage of a doomed or useless limb. Twenty-one limbs were retrospectively
studied to determine which variables influenced salvage or loss.
A limb-salvage score was developed that weighted the level of the vascular
injury, the degree of osseous injury, the degree of muscle injury,
and the warm ischemia time. With use of the same limb cohort to
develop and validate the PSI, Howe et al. reported a sensitivity
of 78% and a specificity of 100%. In the current study, we were
not able to reproduce these findings. The sensitivity and specificity
of the PSI for the patients with an ischemic limb injury were 56%
and 79% when immediate amputations were included in the analysis
and 40% and 79% when immediate amputations were excluded. The performance
was not improved when only open tibial fractures were considered.
Mangled Extremity Severity Score (MESS)
The MESS was proposed by Johansen et al.4 in
1990. Like the PSI, the MESS was designed to address limbs with
combined vascular and orthopaedic injuries. Johansen et al. proposed
that the MESS be used to select lower-extremity injuries that warranted
primary amputation. However, vascular injury was never clearly defined
in the MESS scoring system, and the MESS score allows for evaluation
of patients with normal perfusion2,4.
For this reason, the MESS has been widely referenced as the trauma
limb-salvage index for lower-extremity trauma1,7,21.
The MESS evaluates four characteristics related to injury: degree
of tissue injury, severity of limb ischemia, presence and duration
of shock, and patient's age. The score assumes that the response to
trauma and limb salvage in a patient who is twenty-nine years old
will differ from that in one who is thirty years old and that a
transient depression in blood pressure is clinically important and could
affect the outcome potential of the limb. The MESS was developed
retrospectively in a study of twenty-five patients. The index was
then validated in that same patient group and in a group of twenty-six
additional limbs that were assessed prospectively. Johansen et al.4 concluded that a MESS score of 7
or more was 100% predictive of amputation. The performance of the
MESS in our prospective series did not duplicate these findings.
If all of the limbs in the present study were considered, the sensitivity
of the MESS was 46%. This increased to 72% if only the ischemic
limbs were considered. No advantage was noted in the application
of the MESS to any of the nonischemic-limb subgroups. The sensitivity decreased
to 27% when the immediate amputations were excluded from the analysis.
The sensitivity of the MESS in the cohort of the severely open tibial fractures
was 45% overall but was only 22% when the immediate amputations
were excluded. The area under the curve for the limbs with a type-IIIC tibial
fracture indicated a poor discriminative ability of the index, regardless
of whether the immediate amputations were included or excluded (0.62 and
0.68, respectively). Generally, the MESS was highly specific, suggesting
that it might be useful in predicting the limbs that should not
undergo amputation. The low sensitivity, however, suggests that
a large proportion of limbs eventually requiring amputation would
be at risk for a delay in the procedure, and this delay might in
turn be associated with complications.
Nerve Injury, Ischemia, Soft-Tissue Injury,
Skeletal
Injury, Shock, and Age of Patient Score (NISSSA)
McNamara et al.5 introduced
the NISSSA score, in 1994, to address perceived weaknesses of the
MESS. The authors envisioned an application similar to that of the MESS,
at the time of initial limb evaluation and clinical decision-making.
Specifically, the NISSSA added a nerve-injury component, giving
the highest weight to the loss of plantar sensation, and divided tissue
injury into soft and skeletal variables. Twenty-six limbs were scored
retrospectively with the MESS and NISSSA methods. Compared with the
MESS score, the NISSSA score was found to be more sensitive (81.8%
compared with 63.6%) and more specific (92.3% compared with 69.2%).
Both scores were reported to be highly accurate in predicting amputation.
The present study did not confirm these findings. The NISSSA had
a sensitivity of 33% when applied to all type-III tibial fractures and
of 13% when immediate amputations were excluded. The performance
did not improve when the type-IIIB and IIIC tibial fracture subgroups were
analyzed separately.
Limb Salvage Index (LSI)
The LSI was developed by Russell et al.6,
in 1991, to assist with the decision-making process for limb trauma
associated with vascular injury. Absolute indications for amputation included
a score of 6 or more. Seventy limbs were evaluated retrospectively.
Twenty-six limbs had pulse deficits requiring revascularization.
The LSI includes seven components related to injury: arterial, nerve,
bone, skin, muscle, and deep venous injury as well as warm ischemia
time. A 100% correlation between the limb outcome and the threshold
score was reported6. The present
study did not confirm that finding. When the LSI was applied to
the eighty-eight ischemic limbs, we found a sensitivity of 83% and a
specificity of 82%. The LSI did, however, perform significantly
better than the MESS and the PSI when the scoring systems were applied
to all of the limbs in the study (p < 0.05), and it performed significantly
better than did the MESS, PSI, NISSSA, and HFS-97 when only type-III
tibial fractures were considered (p < 0.05). This difference
might be related to the injury focus of the score, the anatomical
evaluation of the nerve injury, and/or the weight assignment selected
by Russell et al. for each component.
Hannover Fracture Scale (HFS)
The HFS was developed to quantify risk factors for injury and
complications in high-energy trauma to a limb. First reported in
199322, the initial thirteen characteristics
related to injury, their relative score weight, and the recommended amputation
threshold have been refined by a continued reassessment strategy
with use of multiple regression analysis and receiver operator characteristic
curves7. The Hannover group's
recent findings of a high sensitivity and specificity were not confirmed
by the present study. Overall, the HFS-97 had a sensitivity of 37%
(10% for type-IIIB fractures and 67% for type-IIIC fractures). The
sensitivity decreased to 11% when the immediate amputations were excluded
from the analysis. The HFS-97 did perform significantly better than
the MESS and the NISSSA when only the type-IIIC fractures were considered
(p < 0.05). The area under the receiver operating characteristic
curves suggested that the HFS-97 is only moderately good at discriminating between
limbs that will undergo amputation and those that will be salvaged.
Our evaluation of the HFS-97 was limited, however, by the necessity
of modifying the score to account for the clinical practice in the
LEAP institutions of not performing bacteriological studies of specimens
from the initial wound.
Previous studies have challenged the utility of the lower-extremity
injury-severity scores. Bonanni et al.9 retrospectively
studied fifty-eight severely traumatized limbs and their MESS, LSI,
and PSI scores. In addition to limb amputation that was performed during
the initial hospitalization or the two-year follow-up period, selected
functional outcomes were used to define failed attempts at limb
salvage. These included an insensate extremity and a salvaged limb
that impeded the patient's ability to walk at least 150 feet (46
m), climb twelve stairs, transfer from a bed or a tub, or move into
a seated position from a standing position or vice versa. However,
only two patients were considered to have had a functional-outcome
failure, and their inclusion in the final analysis of the sensitivity
and specificity had little impact on the result. Bonanni et al.
found relatively low sensitivities for the indices (the MESS had
a sensitivity of 22%; the LSI, of 61%; and the PSI, of 33%), and
they concluded that the indices had no predictive utility. Durham
et al.10 analyzed the MESS, LSI,
and PSI for fifty-one lower-extremity injuries retrospectively and reported
that the MESS had a sensitivity and specificity of 79% and 83%;
the LSI, of 83% and 83%; and the PSI, of 96% and 50%. They found
no correlation between the long-term function and the severity scores.
The present study did not support the utility of any of the lower-extremity
injury-severity indices for discriminating between the limbs requiring
amputation and those likely to be salvaged successfully. Overall,
the lower-extremity injury-severity scores lack sensitivity, although
in some cases they were very specific. This might suggest that,
while the indices are incapable of identifying patients who will
eventually require an amputation, they might be useful as a screening
test to support the entry of an extremity into the limb-salvage
pathway. They may also be useful in research as a means of adjusting
for severity of a high-energy injury to the lower extremity.
Assessing the performance of any lower-extremity injury-severity
score depends on the characteristics of the sample population. Specifically,
the inclusion or exclusion of immediate amputations significantly
alters the estimates of sensitivity (the specificity of the index,
by definition, will remain the same whether or not immediate amputations
are included in the analysis). Estimates of sensitivity in the present
analysis decreased by as much as 50% to 70% when immediate amputations
were excluded from the analysis (p < 0.05). While one may argue
that a more appropriate and clinically meaningful assessment of
a limb-salvage score would exclude immediate amputations, it is
important to underscore that, with few exceptions, sensitivities
were found to be uniformly low in the present study, even when immediate
amputations were included. Sensitivity of the LSI and MESS reached
modest levels when used to predict amputations among patients with
an ischemic limb (83% and 72%, respectively) and patients with a
type-IIIC tibial fracture (91% and 78%, respectively).
To our knowledge, the present study was the first large, independent,
prospective evaluation of the lower-extremity injury-severity scores.
The strengths of this study include its prospective design, the
well-defined inclusion and exclusion criteria, and the overall size
of the study cohort. Nonetheless, the results should be interpreted
with some caution. First, although the overall size of the sample
was substantial, the small numbers included in some of the subgroup
analyses resulted in unstable estimates of both sensitivity and
specificity, as indicated by wide confidence intervals. In addition, since
all patients were treated at level-I trauma centers with well-established
orthopaedic trauma programs, the results may not be generalizable
to other hospitals. Finally, in this analysis, the scoring methods
were examined for their ability to predict limb viability as defined
by the need for an amputation within six months after discharge.
Of equal if not greater interest is the extent to which the scoring methods
predict long-term functional outcomes or correlate with resource
utilization and complication rates.
The performance of the indices in all of the injury-pattern groups
indicates that these lower-extremity injury-severity scoring systems
have limited usefulness and cannot be used as the sole criterion
by which amputation decisions are made. Scores at or above the amputation
threshold should be used cautiously by a surgeon who must decide
the fate of a lower extremity with a high-energy injury.
Inclusion Criteria
1. Traumatic amputation below the distal aspect of the femur
2. Gustilo type-IIIA tibial fracture with:
(a) a hospital stay of more than four days,
(b) two or more surgical procedures involving the limb, and
(c) two or more of the following: (i) severe muscle damage (loss
of more than 50% of one or more major muscle groups or associated
compartment syndrome with myonecrosis), (ii) associated nerve injury
(posterior tibial or peroneal deficit), (iii) major bone loss or
bone injury (associated with a fibular fracture and displacement
of more than 50%, comminuted and segmental fracture, and more than
75% probability of requiring bone graft/transport)
3. Gustilo type-IIIB tibial fracture
4. Gustilo type-IIIC tibial fracture
5. Dysvascular injuries below the distal aspect of the femur
(not the foot), including dislocations of the knee, closed tibial
fractures, and penetrating wounds with vascular injury noted on
arteriograms or ultrasound or at the time of surgery
6. Major soft-tissue injuries below the distal aspect of the
femur (not the foot), including:
(a) AO/ASIF type-IC3 (circumscribed), IC4 (extensive closed),
and IC5 (necrotic from a contusion) degloving injuries5,
(b) severe soft-tissue crush or avulsion injuries with muscle
disruption or compartment syndrome, and
(c) dompartment syndrome resulting in myonecrosis and requiring
partial or full resection of the muscle-unit
7. Severe injuries to the distal aspect of the tibia or to the
foot, including:
(a) type-III open fractures of the pilon,
(b) type-IIIB open fractures of the ankle, and
(c) severe open injury to the hindfoot or midfoot (insensate
plantar surfaces, devascularization, major degloving injury, or
open soft-tissue injury requiring coverage)
Gregory RT; Gould RJ; Peclet M; Wagner JS; Gilbert DA; Wheeler JR; Snyder SO; Gayle RG; and Schwab CW: The mangled extremity syndrome (M.E.S.): a severity grading
system for multisystem injury of the extremity. J Trauma.,1985.25: 1147-50, 251147
1985
[PubMed]
Helfet DL; Howey T; Sanders R; and Johansen K: Limb salvage versus amputation: preliminary results of
the Mangled Extremity Severity Score. Clin Orthop,1990.256: 80-6, 25680
1990
[PubMed]
Howe HR Jr; Poole GV; Hansen KJ; Clark T; Plonk GW; Koman LA; and Pennell, TC: Salvage of lower extremities following combined orthopedic
and vascular trauma. A predictive salvage index. Am Surg,1987.53: 205-8, 53205
1987
[PubMed]
Johansen K; Daines M; Howey T; Helfet D; and Hansen ST Jr: Objective criteria accurately predict amputation following
lower extremity trauma. J Trauma,1990.30: 568-73, 30568
1990
[PubMed]
McNamara MG; Heckman JD; and Corley EG: Severe open fracture of the lower extremity: a retrospective
evaluation of the Mangled Extremity Severity Score. J Orthop Trauma,1994.8: 81-7, 881
1994
[PubMed]
Russell WL; Sailors DM; Whittle TB; Fisher DF Jr; and Burns RP: Limb salvage versus traumatic amputation. A decision based
on a seven-part predictive index. Ann Surg,1991.213: 473-81, 213473
1991
[PubMed]
Tscherne H. Personal communication,
1999
Tscherne H, and Oestern HJ: A new classification of soft-tissue damage in open and
closed fractures. Unfallheilkunde,1982.85: 111-5, German85111
1982
[PubMed]
Bonanni F; Rhodes M; and Lucke JF: The futility of predictive scoring of mangled lower extremities. J Trauma,1993.34: 99-104, 3499
1993
[PubMed]
Durham RM; Mistry BM; Mazuski JE; Shapiro M; and Jacobs D: Outcome and utility of scoring systems in the management
of the mangled extremity. Am J Surg,1996.172: 569-74, 172569
1996
[PubMed]
Lange RH: Limb reconstruction versus amputation decision making
in massive lower extremity trauma. Clin Orthop,1989.243: 92-9, 24392
1989
[PubMed]
Gustilo RB; Mendoza, RM; and Williams DN: Problems in the management of type III (severe) open fractures:
A new classification of type III open fractures. J Trauma,1984.24: 742-6, 24742
1984
[PubMed]
Baker SP; O'Neill B; Haddon W Jr; and Long WB: The injury severity score: a method for describing patients
with multiple injuries and evaluating emergency care. J Trauma,1974.14: 187-96, 14187
1974
[PubMed]
Orthopaedic Trauma Association
Committee for Coding and Classification: Fracture and dislocation compendium. J Orthop Trauma,1996.10 Suppl 1: 10 Suppl 1
1996
Müller ME, Allgöwer M, Schneider R,
Willenegger H, editors. Manual of internal fixation.
Techniques recommended by the AO-ASIF Group. New York:
Springer; 1991
Zweig MH, and Campbell G: Receiver-operating characteristic (ROC) plots: a fundamental
evaluation tool in clinical medicine. Clin Chem,1993.39: 561-77, 39561
1993
[PubMed]
McDowell I, Newall C. Measuring
health: a guide to rating scales and questionnaires 2nd
ed. New York: Oxford University Press; 1996. p 32
Swets JA: Measuring the accuracy of diagnostics systems. Science,1988.240: 1285-93, 2401285
1988
[PubMed]
DeLong ER; DeLong DM; and Clarke-Pearson DL.: Comparing the areas under two or more correlated receiver
operating characteristic curves: a nonparametric approach. Biometrics,1988.44: 837-45, 44837
1988
[PubMed]
Bondurant FJ; Cotler HB; Buckle R; Miller-Crotchett P; and Browner BD: The medical and economic impact of severely injured lower
extremities. J Trauma,1988.28: 1270-73, 281270
1988
[PubMed]
Behrens FF. Fractures with
soft tissue injuries. In: Browner BD, Jupiter JP, Levine AM, Trafton
PG, editors. Skeletal trauma. Fractures, dislocations, ligamentous
injuries. Volume 1. Philadelphia: WB Saunders; 1998. p 391-418
Suedkamp NP; Barbey N; Veuskens A; Tempka A; Haas NP; Hoffmann R; and Tscherne H.: The incidence of osteitis in open fractures: an analysis
of 948 open fractures (a review of the Hannover experience). J Orthop Trauma,1993.7: 473-82, 7473
1993
[PubMed]