Opinions about the treatment of Legg-Calvé-Perthes disease vary
widely. The earliest treatment programs arose from the methods used to treat
tuberculosis of the hip. Those methods included spica cast immobilization, bed
rest, traction, and walking with a weight-relieving caliper. Surgical methods
were also employed in the early years, with Legg, Calvé, and
Waldenstrom eventually coming to the conclusion that treatment had little
effect on the
outcome1-3.
Some modern authors have recommended nonoperative means such as bracing and
cast immobilization and have reported satisfactory outcomes for most
patients4-9.
Many authors have reported good results with femoral varus
osteotomies10-14.
Others have recommended the Salter innominate osteotomy as an effective
treatment
alternative10,15-18.
Still others have advocated more complex surgical intervention, including
combined
osteotomies19,20,
shelf
arthroplasty21-23,
Sugioka
osteotomy24, and
Chiari
osteotomy25-27.
However, it is difficult to draw firm conclusions from these studies as few
were prospective or controlled, different criteria were used for inclusion,
surgical indications varied considerably, and various age groups were
studied28.
The Legg-Perthes Study Group includes Peter F. Armstrong, John G. Birch,
William H. Browning, Alvin H. Crawford, Peter A. DeLuca, Frederick R. Dietz,
Anne K. Dzus, Keith Gabriel, Neil E. Green, Richard H. Gross, Curtis R. Gruel,
William A. Herndon, John A. Herring, Brian L. Hotchkiss, James G. Jarvis,
Charles E. Johnston II, Vicki Kalen, John D. King, Dayle L. Maples, John B.
McIvor, Peter L. Meehan, Marc J. Moreau, Raymond T. Morrissy, Colin F.
Moseley, Scott J. Mubarak, Richard Nicol, George T. Rab, B. Stephens Richards,
Lawrence A. Rinsky, James W. Roach, J. Andy Sullivan, Elizabeth A. Szalay,
George H. Thompson, O. Timothy Ward, Hugh G. Watts, Stuart L. Weinstein,
Dennis R. Wenger, David A. Yngve, Seymour Zimbler. Diane Ramey was the study
coordinator. The eight pediatric orthopaedists in the Pediatric Orthopaedic
Study Group included Hugh H. Watts, George H. Thompson, George T. Rab, John A.
Herring, W. Timothy Ward, Stuart L. Weinstein, J. Andy Sullivan, and Neil E.
Green.
The present study was initiated following a debate in 1982 at the Pediatric
Orthopaedic Study Group. Eight pediatric orthopaedists had a series of
meetings to develop protocols for a prospective, long-term study of new
patients with Legg-Calvé-Perthes disease. The pediatric orthopaedic
community was informed of the study, and thirty-nine pediatric orthopaedists
committed to and enrolled patients in the Legg-Calvé-Perthes Study
Group. The first patients were enrolled in 1984 and were followed to skeletal
maturity. This is a report of the outcomes in those patients.
The study was designed to answer two fundamental questions. First, do any
of the proposed treatment methods alter the outcome of
Legg-Calvé-Perthes disease? Second, is it possible to identify specific
factors that predict the outcome for patients with this diverse disorder?
Study Design
The study was prospective, with follow-up to skeletal maturity of children
six to twelve years old at the onset of the disease who had had no prior
treatment and in whom the disease was in its initial stages. Hips that had
reached the radiographic stage of reossification at presentation were
excluded. Each investigator agreed to use only one treatment mode, and to
follow the protocol of that method, when managing all patients who met the
study criteria. Children were excluded if they had a hematologic disorder, a
history of steroid therapy, hypothyroidism, juvenile arthritis, diabetes,
renal failure, any metabolic or neoplastic disease, a family history or
radiographic findings of multiple epiphyseal dysplasia, a hip infection, a
history of developmental dysplasia of the hip, or failure to follow protocol.
Radiographs of the knees and spine were required if a patient presented with
bilateral osteonecrosis of the hip with a similar radiographic stage of
evolution in both hips, and the patient was excluded if findings confirmed a
diagnosis of epiphyseal dysplasia. We carefully explained the study to the
families, and informed consent to participate was obtained from the parents of
each participant. The right to withdraw from the study at any time without
penalty was explained as well. The study began prior to the availability of
the institutional review board process.
Treatment Protocols
The patient was seen as often as needed by the treating physician, and data
sheets and radiographs were submitted at the time of induction into the study;
at four, eight, and twelve months during the first year; and yearly
thereafter. The range of motion of the hips was recorded, and standing
anteroposterior and supine frog-leg lateral radiographs of the pelvis were
made at each visit. A wrist radiograph was made to determine bone age on the
first visit. Investigators who failed to submit data were prompted to do so by
mail and telephone. Patients who did not return for visits were contacted, and
their expenses for follow-up examinations and radiographs were reimbursed.
Treatment Groups
No treatment: Patients in this group were not treated with any
specific measures except for treatment of symptoms, including nonsteroidal
anti-inflammatory medications as necessary and reduction of activities at the
parents' discretion.
Range-of-motion treatment: Patients in this group were instructed
to perform an exercise program consisting of standing abduction stretching and
active range-of-motion exercises of the hip at least once a day. Loss of range
of motion was treated with a range-of-motion plan. Initially, active-assisted
range-of-motion exercises of the hip, emphasizing abduction, were prescribed.
If loss of motion was persistent, traction was also used at the investigator's
discretion. The traction was usually applied at home, but if it was applied in
the hospital, it was used for no more than two weeks in that setting. For
persistent loss of motion, adductor releases were performed, usually followed
by application of abduction plaster casts, termed Petrie
casts29.
The casts were used for a maximum of six weeks and for no more than two
episodes. The goal was to achieve =30° of hip abduction as documented
on an anteroposterior radiograph of the pelvis. The patient was taught to walk
while wearing the casts.
Brace treatment: The Atlanta Scottish Rite brace was ordered at
the time of induction into the study. The brace was adjusted to produce
=30° of hip abduction on a standing radiograph. The patient was
instructed to wear the brace at all times except for one hour for bathing and
up to three hours for swimming. The patient was told to avoid weight-bearing
when not wearing the brace. The patient's (or parent's) estimate of the hours
per day that the brace was worn was reported to the study. No patient was
dropped from the study because of noncompliance with brace wear. The
range-of-motion protocol was used for patients in the brace-treatment group
when substantial loss of motion was noted. Brace treatment was discontinued
when radiographs showed evidence of early reossification.
Innominate osteotomy: After the initial evaluation, the patients
in the innominate osteotomy group were either followed for a period of time to
determine the disease severity or were scheduled for surgery at presentation,
at the treating surgeon's discretion. Treatment for loss of motion prior to
the surgery could include the range-of-motion plan with one episode of use of
abduction casts. The osteotomy was performed with use of a 30° to 35°
wedge of bone obtained from the anterior aspect of the pelvis with movement of
the acetabular segment of the pelvis forward 5 to 10 mm. The osteotomy was
closed posteriorly and fixed with two large threaded Steinmann pins. Adductor
tenotomy and iliopsoas recession were recommended. Postoperatively, the use of
a spica cast was optional, and it was not used for more than six weeks.
Partial weight-bearing with use of crutches was allowed for patients who were
not wearing a cast. Fixation pins were removed at the surgeon's
discretion.
Femoral osteotomy: As in the previous group, the surgical
procedure could be planned on the first visit or on a subsequent visit. One
trial of the range-of-motion protocol and use of abduction casts was an
initial option prior to surgery. The femoral osteotomy was done at the level
of the lesser trochanter with blade-plate fixation. The investigators agreed
that a neck-shaft angle of 110° to 115° with little or no rotational
change was desirable. Use of a spica cast for six weeks postoperatively was
optional, and the patients who were not treated with a cast were allowed to
walk with partial weight-bearing. The implant was removed at the surgeon's
discretion.
In each treatment group, a provision was made for a child who had clinical
and radiographic findings indicating a very mild form of
Legg-Calvé-Perthes disease. The criteria for this category included
minimal symptoms, maintenance of a full or nearly full range of motion without
treatment, and radiographic findings of minimal disease. The children in this
group could be observed without formal treatment regardless of the treatment
group in which they had been enrolled. Treatment would be instituted if they
subsequently lost range of motion or if greater radiographic changes
developed.
Clinical Information
The age at the onset of the disease, sex, height, weight, and range of
motion of the hip were noted. At each visit, the range of motion, compliance
with brace wear (when appropriate), complications, and use of Petrie casts
were recorded. Investigators recorded the
Catterall30
classification and the
Salter-Thompson31
radiographic classification of the disease.
Radiographic Measurements
Radiographs were evaluated by one of us (J.A.H.) on receipt at the study
headquarters. Initially, the Catterall classification was used, but it was
found to be inconsistent among the investigators even following instructional
meetings. Subsequently, the lateral pillar classification was applied, with
use of the criteria described in Part I, which precedes this article in this
issue of The Journal. Skeletal age was determined from radiographs of
the hand and wrist with use of the Greulich and Pyle
atlas32. When no
radiograph of the wrist was available, the skeletal age was determined from
the initial pelvic radiograph with use of the Oxford bone age
method33. An Oxford
score of 18 was considered equivalent to a wrist-radiograph-derived skeletal
age of six years. We rated the final radiographs with use of the Stulberg
classification, as described in Part I of this article. In the femoral
osteotomy group, the initial postoperative and final neck-shaft angles were
measured and the final articular-trochanteric distance was measured in
millimeters. The Waldenstrom
stage3 at the time
of surgical treatment was recorded.
Exclusions
Exclusion forms submitted: There were 459 exclusion forms
submitted for patients who were never enrolled in the study. Of these
patients, 277 were younger than six years of age, twelve were more than twelve
years of age, nine had had prior surgery, six had had steroid treatment, one
had had a prior infection, three had a traumatic injury or developmental
dysplasia of the hip, the hip was already healing at presentation in
forty-four, the parents of ten denied consent, two had hemoglobinopathies,
five had metabolic or neoplastic disease, five had epiphyseal dysplasia,
thirty-four were seen for a second opinion but were not treated, and fifty-one
had miscellaneous other reasons for exclusion.
Attrition: In the brace and range-of-motion treatment groups, nine
hips were excluded from the analysis because they had had surgical
intervention prior to the healing phase of the disease. Five of these hips had
a varus femoral osteotomy, and four had an innominate osteotomy. Three hips in
the femoral osteotomy group were excluded: one because a valgus osteotomy had
been performed, one because a triple pelvic osteotomy had been done, and one
because no surgery had been carried out. In addition, four hips in the
innominate osteotomy group were excluded because no surgery had been carried
out.
Initially, 438 patients with 451 affected hips were enrolled. Eighty-five
of these hips were lost to follow-up prior to skeletal maturity, one patient
(two hips) withdrew from the study, one patient (two hips) had Gaucher
disease, and one patient (one hip) died in an automobile accident prior to
skeletal maturity. There was no significant difference between those lost to
follow-up and the study group relative to the age at the onset of the disease,
gender, lateral pillar classification, or treatment group. This left 345 hips
in 337 patients (271 boys and sixty-six girls with 162 right and 183 left
hips) at skeletal maturity. Of the twenty-eight patients with bilateral
disease, eight had both hips included in the study and twenty had one hip
included in the study. The average age at the time of induction into the study
was 7.98 years old (range, 6.0 to 12.0 years old) in the series as a whole,
8.28 years old (range, 6.0 to 12.0 years old) for the girls, and 7.91 years
old (range, 6.0 to 11.7 years old) for the boys. The average skeletal age was
6.26 years old (range, 3.0 to 13.0 years old) in the series as a whole, 7.28
years old (range, 3.0 to 13.0 years old) for the girls, and 6.01 years old
(range, 3.0 to 11.5 years old) for the boys. The mean bone-age delay at
induction was 1.0 years for the girls and 1.9 years for the boys. The average
duration of follow-up was 8.6 years (range, seventeen months to 15.3 years).
All patients were followed for more than two years, except for one girl who
reached skeletal maturity at seventeen months after the onset of the disease
and had no more follow-up. The average age at the time of final follow-up was
16.6 years (range, 9.4 to twenty-three years).
There were nineteen hips in the no-treatment group, seventy-seven in the
range-of-motion group, 129 in the brace group, fifty-two in the femoral
osteotomy group, and sixty-eight in the innominate osteotomy group.
Statistical Methods
Data presented in two-way contingency tables were analyzed with standard
chi-square methods, unless the sample size was judged to be unacceptably
small. If the expected value of any cell in the contingency table was <3.0,
the Fisher exact test was used. We used the SAS statistical program (version
8.2; SAS Institute, Cary, North Carolina) for all calculations. To be
considered significant, the associated p value had to be = 0.05.
If the contingency table compared two or more groups that had more than two
possible responses, the chi-square method was used to compare the groups to
assess whether they had the same pattern of percentages across the responses
seen.
For a total sample size of 345, a contingency table with two rows and three
columns (one commonly seen in the Results section) would have about 80% power
for an alpha of 0.05, if the noncentrality parameter is =0.0286 with two
degrees of freedom. There are many patterns of outcomes that would meet that
criterion. One example is one in which the rows represent two treatments and
the columns represent three outcomes. The overall percentages of hips with
Stulberg classes I or II, III, and IV or V outcomes in this study were roughly
50%, 35%, and 15%, respectively. If the first treatment has outcome rates of
50%, 35%, and 15%, respectively, and the second treatment has outcome rates of
60%, 20%, and 20%, respectively, the chi-square test has an 81% chance of
yielding a significant result. For a total number of 100, the second treatment
would have to have a pattern such as 75%, 10%, and 15% to have 80% power. For
a total number of fifty, the second treatment would have to have a pattern
such as 90%, 5%, and 5%.
In order to maximize the power of the statistical tests, no Bonferroni
correction was applied to the p
values34. We
understand that this decision increases the chance of a Type-I error in any
declaration of significance. When the sample size for a comparison was
relatively small, a finding of no significant difference should not be
interpreted as meaning that there was no difference, but only that we were
unable to detect it with the given sample size (i.e., a Type-II error).
A multinomial logistic regression analysis was used to determine which
combination of factors affected the probability of a particular outcome as
classified with the system of Stulberg et al. The three outcomes were Stulberg
class I or II, Stulberg class III, and Stulberg class IV or V. The factors
that were considered included treatment group (no, range-of-motion, brace, or
surgical treatment), age at induction into the study (8.0 years old or less
and more than 8.0 years old), gender, and lateral pillar classification. The
six first-order interaction terms of these factors were incorporated into the
model to determine if, for example, the effect of treatment on outcome
depended on the gender of the subject. To eliminate the nonsignificant
factors, a backward elimination process was used. The software used for the
analysis was the LOGISTIC procedure in SAS (version 8.2; SAS Institute), with
a p value of =0.05 required for a factor to remain in the model. The Wald
chi-square values were used as a gross measure of the relative importance of a
factor or the interaction of factors. The odds ratios and their associated
confidence intervals were calculated from the resulting regression model.
Results Related to Treatment
In the entire study, twenty-three results were Stulberg class I, 153 were
class II, 116 were class III, fifty-one were class IV, and two were class V
(Table I). Stulberg class-I and
II results are combined for the remainder of this report, as they have the
same benign natural history. Stulberg class-IV and V results are also combined
because of the rarity of the class-V outcome. The five treatment groups were
not significantly different with respect to age at the onset of the disease or
gender. The distribution of the lateral pillar classifications within the
treatment groups was similar except that there were fewer lateral pillar
group-C hips in the innominate osteotomy group than in the other treatment
groups. Petrie casts were used in all groups except the no-treatment group. In
the brace group, thirty-two hips were treated with Petrie casts; one patient
had two applications, and one had three. In the femoral osteotomy group,
twenty hips were treated with Petrie casts; no patient had two applications
and one had three. The two cases with three applications of Petrie casts were
considered as minor protocol deviations and remained in the study. In the
range-of-motion group, twenty-nine hips were treated with Petrie casts and
three patients had two applications. In the innominate osteotomy group,
nineteen hips were treated with Petrie casts, and one had two
applications.
Femoral osteotomy was performed during the stage of increased density in
forty-four hips, during the early fragmentation stage in six, and during the
late fragmentation stage in two. Innominate osteotomy was performed during the
stage of increased density in forty-nine hips, during the early fragmentation
stage in sixteen, and during the late fragmentation stage in three.
There were no significant differences in outcome among the hips in the
brace treatment group, the no-treatment group, and the range-of-motion group
(p = 0.13) (Table I). Given
that there were no significant differences among these groups and the fact
that all three of the treatment methods are nonoperative, we chose to combine
them into one nonoperative group, thereby increasing our sample size for
analysis. There were also no significant differences in outcome between the
hips in the femoral osteotomy group and those in the innominate osteotomy
group (p = 0.65), and we combined those two treatment groups into one surgical
treatment group (Table I).
The outcome distribution in the combined surgical group was significantly
better than that in the combined nonoperative group (p = 0.02). In the
combined surgical group, there were seventy-two (61%) Stulberg class-I or II,
thirty-five (29%) Stulberg class-III, and twelve (10%) Stulberg class-IV or V
results. In the combined nonoperative group, there were 104 (46%) Stulberg
class-I or II, eighty-one (36%) Stulberg class-III, and forty-one (18%)
Stulberg class-IV or V results (Table
II). The better outcomes in the surgical group were due to better
outcomes in the children in the lateral pillar B group and B/C border group
who had the onset of the disease after their eighth birthday
(Table II). In group B,
twenty-four (73%) of the thirty-three operatively treated hips in the older
children had a Stulberg class-I or II result compared with twenty-seven (44%)
of the sixty-two nonoperatively treated hips in the older children (p =
0.02).
In the children over the age of 8.0 years at the onset of the disease in
the B/C border group, the outcomes after operative treatment were also better
than those after nonoperative treatment (p = 0.05). There were few Stulberg
class-I or II results in either treatment group, and there was a higher
frequency of Stulberg class-III results in the operative group and a greater
percentage of Stulberg class-IV or V results in the nonoperative group.
Younger children (8.0 years of age or less at the onset of the disease) in
lateral pillar group B had very good results after either operative or
nonoperative treatment, with >75% having a Stulberg class-I or II outcome.
Good results were uncommon in lateral pillar group C, and, with the numbers
available, no significant differences in outcome were noted between the
operative and nonoperative treatment groups or among specific ages of disease
onset (Table II).
Similar results were found in comparisons based on skeletal age at the
onset of the disease (Table
III). In the lateral pillar B group and B/C border group, children
with a skeletal age of greater than six years at the onset of the disease had
significantly better outcomes after operative treatment than after
nonoperative treatment (p = 0.007 for group B and p = 0.04 for the B/C border
group). In children with a bone age of six years or less at the onset of the
disease, there was no significant difference between the outcomes of operative
and nonoperative treatment (p = 0.84 and 0.28). Lateral pillar group-B hips in
patients with a bone age of six years or less generally did well, and no
significant treatment effect was found (p = 0.84). The results in lateral
pillar group C were frequently poor and were similar in the operative and
nonoperative groups overall (p = 0.20) and in the younger (p = 0.51) and older
(p = 0.26) cohorts.
Factors Other Than Treatment Associated with Outcome
The age at the onset of the disease was highly correlated with outcome
(Table IV). Children in whom
the disease began at the chronologic age of 8.0 years or less had
significantly better results across all the treatment groups than did children
older than 8.0 years (p = 0.0001). Children with a bone age of six years or
less at the onset of the disease also had better outcomes than did those with
a bone age of more than six years (p = 0.0002).
The lateral pillar classification had an even stronger correlation with
outcome than did the age at the onset of the disease
(Table V). All lateral pillar
group-A hips had a Stulberg class-I or II outcome. In lateral pillar group B,
145 (67%) of the 218 hips had a Stulberg class-I or II outcome whereas only
fourteen (6%) of the 218 had a Stulberg class-IV or V result. The results for
the hips in the lateral pillar B/C border group were intermediate between
those in groups B and C. The hips in lateral pillar group C had the worst
results: of the sixty hips, only eight (13%) had a Stulberg class-I or II
result, thirty-one (52%) had a Stulberg class-III result, and twenty-one (35%)
had a Stulberg class-IV or V result. The difference in outcome among the
lateral pillar groups was significant (p < 0.0001).
Considered together, the age at the onset of the disease and the lateral
pillar classification were highly predictive of outcome
(Fig. 1). Lateral pillar
group-B hips did well in children of all ages, and the effect of age at the
onset of the disease was not great. In group B, thirty-one (76%) of the
forty-one hips in children who were between 8.0 and 8.9 years old at the onset
of the disease had a Stulberg class-I or II result, whereas eight (62%) of the
thirteen hips in children who were between 10.0 and 10.9 years old at the
onset of the disease had a Stulberg class-I or II result
(Fig. 1). Stulberg class-I or
II results were found less often in the B/C border group, and there was a
strong decrease in the prevalence of that result with increasing age at the
onset of the disease. (Seven [54%] of the thirteen children who were 6.0 to
6.9 years old at the onset of the disease had a Stulberg class-I or II result.
This rate dropped to only three [19%] of sixteen in the age-group of 8.0 to
8.9 years.) In lateral pillar group C, there were few Stulberg class-I or II
results at any age, with such a result found in five (20%) of the twenty-five
hips in children 6.0 to 6.9 years old, two (13%) of the sixteen in children
7.0 to 7.9 years old, and only one (11%) of the nine in children 8.0 to 8.9
years old. There were no Stulberg class-I or II results in children who were
nine years of age or older at the onset of the disease in group C.
Stulberg class-IV or V results were uncommon and their prevalence increased
slightly with age in lateral pillar group B
(Fig. 2). In group B, only five
(3%) of the 159 hips in children under the age of eight years at the onset of
the disease had a Stulberg class-IV or V result, five (13%) of the thirty-nine
in children between the ages of 9.0 and 9.9 years at the onset of the disease
had a Stulberg class-IV or V outcome, one of the fourteen in children between
the ages of 10.0 and 10.9 years had that result, and four of the seven in
children eleven years of age or older had that result. Stulberg class-IV or V
results occurred more frequently in the B/C border group and in group C, with
an increase in frequency in older children. In the age group of 6.0 to 7.9
years at the onset of the disease, Stulberg class-IV and V results occurred in
four (13%) of thirty B/C-border-group hips and eight (20%) of forty-one
group-C hips. In the age group of nine years and older, eight of fourteen
B/C-border-group hips and eight of ten group-C hips had Stulberg class-IV or V
results.
Girls were more likely to have worse outcomes than boys
(Table VI). Girls had
twenty-six (38%) Stulberg class-I or II, twenty-four (35%) Stulberg class-III,
and nineteen (28%) Stulberg class-IV or V results. Boys had 150 (54%) Stulberg
class-I or II, ninety-two (33%) Stulberg class-III, and thirty-four (12%)
Stulberg class-IV or V results (p = 0.0034). Girls over the chronologic age of
8.0 years at the onset of the disease had worse results than boys in that
age-group (p = 0.004), whereas boys and girls 8.0 years old or younger at the
onset of the disease had similar results (p = 0.77)
(Table VI). The lateral pillar
classifications were similar for the boys and girls (p = 0.96), as was the
distribution of boys and girls among the treatment groups (p = 0.80).
The girls in the study had a 1.0-year delay in bone age compared with a
1.9-year delay for the boys. Girls with a skeletal age of more than six years
at the onset of the disease had worse outcomes than boys of the same skeletal
age (p = 0.03). Girls and boys with younger skeletal ages had similar outcomes
(p = 0.86).
Logistic Regression Analysis
The logistic regression analysis showed that each independent
factor—age at the onset of the disease, treatment, gender, and lateral
pillar classification—was a significant contributor to the Stulberg
outcome. In this model, the lateral pillar classification was the dominant
predictor of outcome, with age at onset next in importance. None of the six
interaction terms formed by the four factors were significant predictors of
the Stulberg outcome. The residual chi-square had a p value of 0.71,
indicating that the final model provided a satisfactory fit to the data. The
Wald chi-square values provide a crude indication of the relative importance
of each factor. The values were 83.3 for the lateral pillar classification,
34.3 for the age at the onset of the disease, 13.8 for the type of treatment,
and 10.1 for gender. The magnitudes of the odds ratios also show that the
lateral pillar classification and age at the onset of the disease are major
factors in determining the Stulberg outcome
(Table VII).
Complications and Later Operations
Complications in the osteotomy group included three cases of pin migration
requiring either removal or replacement, loss of position of the osteotomy
site with posterior displacement of the distal portion in one hip, and three
superficial pin-track infections. In the femoral osteotomy group,
complications included one case of protrusion of the fixation blade through
the femoral neck without an adverse effect. One patient had a fracture distal
to the fixation plate of a femoral osteotomy. Blade-plates were routinely
removed at various time-periods after healing of the osteotomy site, with only
six hips having plates still in place at maturity. All fixation pins in the
innominate osteotomy group were removed.
Other Radiographic Measures
The neck-shaft angles in the femoral osteotomy group ranged from 79° to
130° and averaged 116° on the initial postoperative radiographs and
ranged from 82° to 140° and averaged 124° on the most recent
radiographs. The most recent articular-trochanteric distances in the femoral
osteotomy group averaged 6.8 mm and ranged from -12 to 29 mm.
While there are many reports of treatment of Legg-Calvé-Perthes
disease4-28,
there have been few controlled studies and, to our knowledge, only one
randomized study. Our group initially considered performing a randomized
study, but the prevailing treatment preferences of the members were such that
no agreement on randomization could be reached. Our study, termed a
"best effort" study, allowed the participating surgeon to select
one treatment for all of his or her patients who met the study criteria. This
design offers many of the advantages of the randomized surgeon design
described by Rudicel and
Esdaile35. Our goal
was to eliminate patient selection bias by requiring the surgeon to carry out
the same predetermined treatment plan for each patient. As an additional
safeguard against selection bias, the surgeons were required to submit
exclusion forms for all patients whom they had seen but had not entered into
the study. It remains possible that other sources of bias may have affected
which patients entered a treatment group; these could include referral pattern
bias, geographic bias, and others.
There are differences of opinion regarding the meaning of the various
Stulberg classes of outcome. Some consider Stulberg class-I, II, and III
results to be acceptable or
good36. Others
consider only class-I and II outcomes to be
good37,38.
We chose to combine Stulberg class-I and II results in our report, and we
consider those to be good outcomes. Stulberg class-III results were considered
to be intermediate, and Stulberg classes IV and V, which we combined, were
considered to be poor results. The reader remains free to interpret the
meaning of these categories.
We chose to combine the nonoperative results and to compare them with the
operative results for several reasons. First, a comparison of the separate
treatment groups subdivided on the basis of known risk factors would have
reduced the sample sizes to a level that made significant differences
difficult to detect. Second, no clinically important differences were found
among the various nonoperative methods or between the two surgical
methods.
A variety of braces have been used to treat Legg-Calvé-Perthes
disease with varying reports of
success29,39-45.
The most widely used brace in the last several decades is the Atlanta Scottish
Rite brace8. While
the authors of early reports found it to be successful, more recent studies
did not show a positive effect on
outcome46-48.
We did not find any improvement in femoral head sphericity in patients
treated with the Atlanta brace. The effectiveness of brace treatment is
difficult to determine in any study, such as ours, in which there is no
accurate measure of compliance. An additional difficulty is the probability
that the hip is not in a so-called contained position after muscle spasm and
loss of range of motion occur. In our study, an initial radiograph with the
patient wearing the brace was required to ensure an adequate degree of
abduction. However, subsequent radiographs were often made with the brace off,
and the hip position in the brace could not be determined. In spite of these
limitations, we believe that our study adds to the evidence that bracing is
not effective in the management of Legg-Calvé-Perthes disease.
We also did not find a positive effect of any of the other nonoperative
measures used in this study. The range-of-motion protocol called for exercises
done at home and periodic use of traction and Petrie casts to regain motion
when necessary. Petrie casts were used for 38% of the hips treated with the
range-of-motion protocol. Compliance with exercises was not measured in this
study and undoubtedly varied widely. The small no-treatment group had no
specific therapeutic measures, and the results in that group were similar to
those in the brace and range-of-motion groups across all ages and degrees of
disease severity at presentation.
We found that surgical treatment, either femoral varus osteotomy or
innominate osteotomy, was significantly correlated with an improved outcome in
children with an onset of the disease after their eighth birthday in the
lateral pillar B group and B/C border group. All hips in lateral pillar group
A and younger children in group B did well overall, and surgical treatment was
not associated with an improved outcome. Hips in group C had no improvement in
outcome with surgical treatment. In addition, there were no significant
differences between the outcomes in the innominate osteotomy and femoral
osteotomy groups. In the vast majority of cases, surgical treatment was
performed in the early stages of the disease, either in the stage of increased
density or in the fragmentation stage. So few hips were treated in the later
part of the fragmentation stage that comparison with those treated earlier was
not possible.
The strongest predictor of outcome was the revised lateral pillar
classification. Stated simply, all lateral pillar group-A hips had a good
result, two-thirds of lateral pillar group-B hips had a good result, almost
half of the B/C-border-group hips had an intermediate result but only one in
four had a good result, and only one in eight lateral pillar group-C hips had
a good result and more than a third had a poor result. These findings support
those of others who noted that the lateral pillar classification is useful in
predicting disease severity and
prognosis49-54.
The second strongest predictor of outcome was age at the onset of the
disease. Children presenting on or before their eighth birthday had a 59% rate
of good results and only an 8% rate of poor results. On the other hand, those
presenting after their eighth birthday had a 39% rate of good outcomes and a
26% rate of poor outcomes. Many authors have confirmed the negative effect of
an older age at the onset of the disease on
outcome7,14,55-66.
An age of less than six years at the onset of the disease is usually, but not
always, associated with a benign course and a good outcome. This study
excluded children of that age, and thus we cannot contribute to the discussion
of appropriate treatment of those children. Children between the ages of six
and eight years at the onset of the disease are considered to have a variable
prognosis and are thought to be likely to benefit from treatment, according to
most
reports12,52,56,62.
Children over the age of nine years at the onset of the disease have fared
less well in response to various therapeutic approaches and are the most
difficult group to manage.
In our study, the combination of chronologic age at the onset of the
disease and the lateral pillar classification was a very strong predictor of
outcome. Within lateral pillar group B, increasing age at the onset of the
disease correlated with fewer hips with a Stulberg class-I or II outcome and
more with a Stulberg class-III result, with only a small increase in the
number of hips with a Stulberg class-IV or V outcome. In the B/C border group,
there was a marked increase in the prevalence of Stulberg class-IV or V
outcomes with increasing age (over eight years old). In lateral pillar group
C, a similar increase in the prevalence of Stulberg class-IV or V outcomes was
observed in children who were eight years of age or older at the onset of the
disease. In groups B/C and C, sixteen (67%) of the twenty-four hips in
patients nine years of age or older had a Stulberg class-IV or V result. In
lateral pillar group B, only ten (17%) of the sixty hips in the same age group
had a Stulberg class-IV or V outcome.
Gender also appears to be a factor in outcome. Girls had less retardation
of their bone age than did boys, which accounts for some of this difference.
However, when the outcomes were compared by bone age, there remained a
significant difference between boys and girls. Thirty-one percent of the girls
with a bone age of more than six years had a Stulberg class-I or II result
compared with 45% of the boys, and 35% of the girls with this bone age had a
Stulberg class-IV or V result compared with 17% of the boys.
The issue of gender influence has been addressed by other
authors7,56,67-69.
Guille et al. compared the initial classification and outcome between 105
girls and 470
boys70. While the
girls were noted to have earlier closure of the capital femoral physis, Guille
et al. did not find a difference in the Catterall or lateral pillar
classification. They did not analyze their treatment groups separately, and
the possibility of a treatment effect cannot be excluded. Lovell et al. found
that the poorer outcome in girls was due to an older age and a more severe
radiographic classification at the onset of the
disease68.
Mukherjee and Fabry found a significant difference in outcome between genders,
with 48% of the girls having a poor result compared with 20% of the
boys69. Those
authors also failed to take treatment effects into consideration.
Several treatment recommendations were substantiated by our study. We found
lateral pillar group-A hips to have an excellent prognosis and to require no
specific treatment. Likewise, children with a lateral pillar group-B hip who
have the onset of the disease on or before their eighth birthday have a good
prognosis and require only symptomatic treatment. Children who have the onset
of the disease on or before their eighth birthday and have a B/C-border-group
hip have a somewhat poorer prognosis but do not appear to benefit from
surgical treatment. Lateral pillar group-B and B/C-border-group hips in
children who are older than 8.0 years of age at the onset of the disease have
significant improvement in outcome if treated with an innominate osteotomy or
a varus femoral osteotomy. Treatment recommendations may also be based on bone
age. Children with a bone age of six years or less at the onset of the disease
do not benefit from surgical treatment. Children with a bone age of greater
than six years and a lateral pillar group-B or B/C-border-group classification
of the hip show a significant benefit from surgical treatment compared with
nonoperative treatment. Children in lateral pillar group C have no evident
improvement with surgical treatment regardless of age, and they have an
unfavorable prognosis. We need to find effective treatment alternatives for
this group. On the basis of these results, it appears that only certain
children with the onset of the disease after the age of 8.0 years or with a
bone age of greater than 6.0 years benefit from surgical treatment with varus
femoral osteotomy or innominate osteotomy.
How early should treatment be initiated? To follow our treatment
recommendations, treatment should begin after the lateral pillar
classification is determined. It is usually possible to determine this
classification within six months after the onset of the disease. On the other
hand, there are advantages to initiating surgical treatment in the early
stages of the disease, before there is substantial femoral head deformity,
which may limit hip mobility. In the great majority of patients in this
series, surgery was performed in the early stages of the disease, often prior
to the fragmentation stage. Our findings support the efficacy of early surgery
for appropriately selected patients.
This study clearly documents risk variables and aids in prognostication,
and we believe that it provides treatment guidelines. Our review of a large
number of radiographs illustrates the wide variability of manifestations of
this disorder and its response to treatment. This variability makes it
necessary to evaluate large numbers of hips to determine treatment efficacy.
We believe that smaller studies can easily lead to incorrect conclusions. At
the same time, however, this variability emphasizes the need for
individualized treatment. We must constantly look for other prognostic factors
and, more importantly, seek treatment approaches that are likely to improve
the outcome for patients with lateral pillar group-C hips, which are most at
risk for a poor result.
The study has several limitations. It was an outcome study based on
radiographic findings. We had considerable range-of-motion data from the first
year of treatment but relatively little information about the clinical status
of the patients thereafter. While it is generally accepted that children with
round femoral heads function well with few symptoms, more specific functional
issues cannot be resolved by this study. Specific questions about the relative
merits, with regard to functional outcome, of femoral osteotomy and the
innominate osteotomy also cannot be answered by this study.
Another limitation was the loss to follow-up of a number of patients. A
great deal of effort was expended on finding and obtaining radiographs of
"missing" patients, and many were found. Some patients were
located but failed repeatedly to keep appointments. The geographic spread of
the study and the mobility of the population contributed to this difficulty.
In addition, the small sample sizes in several of the comparisons raise the
possibility that we missed significant treatment effects that would have been
seen with larger numbers of patients.