ORIGINAL ARTICLE
Operative Versus Nonoperative Treatments for
Legg-Calvé-Perthes Disease: A Meta-Analysis
Nhu-An T. Nguyen, BS,* Guy Klein, DO,w Godwin Dogbey, PhD,* Jessica B. McCourt, PA-C,z
and Charles T. Mehlman, DO, MPHz
Background: Legg-Calvé-Perthes disease (LCPD) is an idiopathic
avascular necrosis of the femoral head primarily affecting children
of ages 4 to 12 years. There is no clear consensus on nonoperative
or operative treatment protocols for pediatric patients presenting
with LCPD. This study uses meta-analysis and a binary logistic
regression model to analyze the radiographic outcomes of these
treatment modalities in pediatric patients.
Methods: Clinical studies describing patients undergoing either
nonoperative or operative treatment of LCPD published from
1960 through 2010 were searched electronically and manually.
Eligible studies consisted of (1) a minimum of 10 patients; (2)
listed age at the time of diagnosis or treatment; (3) performed an
initial severity assessment using the Herring or Catterall classi-
fication; (4) detailed the type of intervention; and (5) reassess-
ment of radiographic outcome after a minimum of 1 year after
treatment using the Mose or Stulberg classification.
Results: Twenty-three studies, 1232 patients, and 1266 hips met
the inclusion criteria. Among patients younger than 6 years, op-
erative and nonoperative treatments are equally as likely to
results in a successful radiographic outcome [odds ratio (OR) =
1.071; P = 0.828; 95% confidence interval (CI), 7.377-32.937]. In
patients older than 6 years, operative treatment is nearly twice as
likely to result in a successful radiographic outcome (OR = 1.754;
P < 0.0001; 95% CI, 1.299-2.370). For age at treatment less than
6 years, a patient treated with a pelvic rather than femoral
procedure was approximately 5 times as likely to have a good
radiographic outcome (w2 = 4.488; P = 0.034; unadjusted
OR = 5.20; 95% CI, 1.021-26.471). Among patients ages 6 or
older, pelvic procedures were equally as likely as femoral proce-
dures to yield a successful radiographic outcome (w2 = 1.845;
P = 0.174; unadjusted OR = 1.329; 95% CI, 0.881-2.004). Sex
From the *Ohio University College of Osteopathic Medicine, Athens;
wOrthopaedic Surgery Residency Program, University Hospitals,
Richmond Heights; and zDepartment of Pediatric Orthopaedic
Surgery, Cincinnati Children’s Hospital Medical Center, Cincinnati,
OH.
Supported in part by the University of Cincinnati Orthopaedic Research
and Education Fund and Division of Pediatric Orthopaedic Surgery
at Cincinnati Children’s Hospital Medical Center, Cincinnati, OH.
The authors declare no conflict of interest.
Reprints: Charles T. Mehlman, DO, MPH, Department of Pediatric
Orthopaedic Surgery, 45229 Cincinnati Children’s Hospital Medical
Center, 3333 Burnet Avenue MLC 2017, 45229 Cincinnati, OH.
E-mail: charles.mehlman@cchmc.org.
Copyright r 2012 by Lippincott Williams & Wilkins
J Pediatr Orthop ! Volume 32, Number 7, October/November 2012
had no significant influence on radiographic outcome (OR =
1.248; P = 0.486; 95% CI, 0.670-2.325).
Conclusions: This meta-analysis suggests that operative treat-
ment is more likely to yield a spherical congruent femoral head
than nonoperative methods among patients 6 years or older.
Among patients younger than 6 years, operative and non-
operative methods have the same likelihood to yield a good
outcome. Patients who were 6 years or older were treated op-
eratively, and had the same likelihood of a good radiographic
outcome regardless of treatment with femoral or pelvic proce-
dures. Among patients younger than 6 years, pelvic procedures
were more likely to result in a good radiographic outcome than
femoral procedures.
Level of Evidence: Level IV meta-analysis.
Key Words: Perthes, operative, nonoperative
(J Pediatr Orthop 2012;32:697–705)
L egg-Calvé-Perthes disease (LCPD) is an idiopathic
pediatric hip disorder associated with avascular ne-
crosis of the femoral head. The resulting hip deformity
may lead to problems ranging from mild limp to severe
debilitating arthritis. LCPD occurs most commonly
among children of 4 to 12 years of age with 4 to 5 times
more males than females being affected.1 Although there
are fewer female cases, some studies have found a worse
prognosis in females presumably because their bones have
less time to grow and remodel as a result of early puberty
compared with their male peers.2–5 The potential influ-
ence of sex on outcome is one of many unanswered
questions related to LCPD.
LCPD is the most common form of pediatric hip
disease. The annual incidence of LCPD cases range from
2 to 294 per 1,000,000 children.6 There is no single ex-
planation for the wide range of incidence of LCPD
cases; however, there is similarity in clinical presentation
among children afflicted with this disease. Some of the
possible explanations given for the wide incidence include
genetic predisposition, environmental exposures, and so-
cioeconomic factors.6 The most noticeable symptom in a
LCPD patient is a mildly discomforting limp, which can
become progressively more painful with increasing phys-
ical activity. The hip joint mobility is restricted in ab-
duction and internal rotation.7 A diagnosis of LCPD is
made with a lateral and anteroposterior x-ray view of the
hip and femoral head. Initial severity of the disease is
www.pedorthopaedics.com | 697
Nguyen et al J Pediatr Orthop
assessed by radiographic methods using the Catterall or
the Herring system involving necrosis in the femoral head,
whereas final assessment is graded with the Mose or
Stulberg classification.
Although discovered over 100 years ago, the precise
etiology of LCPD remains unclear. In some cases, the
disease has the capacity to resolve on its own, but for
many other patients the long-term prognosis may be
substantially worse because of residual hip deformities.
Many patients develop osteoarthritis as early as their
mid-40s.1,8–10 As a result, multiple nonoperative and op-
erative treatment methods have been performed with the
goal of altering the natural history of the disease.
Multiple studies have evaluated nonoperative and
operative treatments with variable conclusions. This lack
of consensus stems from the small sample size and vari-
ability between studies in the age of LCPD study pop-
ulation and severity of femoral head involvement upon
presentation. In addition, the lack of intraobserver or
interobserver reliability in classifying radiographic stages
has weakened the ability to draw consensus opinions
from the LCPD literature.11–13 The primary aim of this
meta-analysis was to compare outcomes of nonoperative
and operative treatments for LCPD. An additional goal
was to determine if long-term prognosis was predicted by
age, sex, or disease severity at initial presentation.
METHODS
Search Strategy
A literature search was conducted using PubMed
MEDLINE (1960 to August 2010), Science Citation
Index (1965 to August 2010), and The Cochrane Library
databases (1960 to August 2010) aimed at identifying
studies examining LCPD treatment results. The following
terms were used: “Perthes” OR “Coxa Plana” and
“treatment” OR “non-operative” OR “conservative” OR
“surgery.” Only full-text articles published in the English
language were considered. We searched for any current or
past clinical trials on LCPD on http://www.clinicaltrials.gov,
current controlled trials (http://www.controlled-trials.com),
http://www.asiaclinicaltrials.org, and the World Health
Organization portal of clinical trials (http://www.who.int/
trialsearch). In addition, abstracts from the following
annual conferences were searched: European Paediatric
Orthopaedic Society (2006 to 2010), Pediatric Orthopae-
dic Society of North America (2007 to 2010), and
American Academy of Orthopaedic Surgeons (2008 to
2010). Bibliographies of 3 major pediatric orthopaedic
textbooks were screened for relevant studies identified
through chapters related to LCPD (Pediatric Orthopedic
Deformities by Frederic Shapiro, Lovell and Winter’s
Pediatric Orthopaedics by Morrissy and Weinstein, and
Tachdjian’s Pediatric Orthopaedics by Herring). Titles and
abstracts that were deemed relevant to this review, were
retrieved as full-text articles to be considered in the
analysis. Authors were contacted in instances when ad-
ditional data were needed. References from selected full-
698 | www.pedorthopaedics.com
! Volume 32, Number 7, October/November 2012
text articles were reviewed to identify additional studies
that potentially met the inclusion criteria.
Study Selection
Inclusion Criteria
Studies that met the inclusion criteria for the meta-
analysis had the following characteristics: (1) a minimum
of 10 patients; (2) initial radiographs documenting se-
verity at the time of diagnosis using the Catterall or
Herring classification; (3) a minimum of 1-year radio-
graphic follow-up using the Mose or Stulberg criteria; (4)
age of patients listed at the time of diagnosis or at the
time of treatment; and (5) the type of intervention used.
Exclusion Criteria
Excluded studies from our analysis had the follow-
ing characteristics: (1) the outcomes of predicted meas-
ures were not fully discernable and data were impossible
to pool and analyze; (2) the combination of Z2 operative
techniques not conventionally used but only reserved as
salvage technique; and/or (3) the patient population
consisted of other hip diseases not limited to LCPD.
Data Extraction
Two reviewers (N.T.N. and J.M.) independently
assessed the level of evidence of the included studies using
the guidelines of the Oxford Centre for Evidence Based
Medicine to determine the level of evidence of the study.
Data that were independently extracted from the selected
studies included: number of patients, number of affected
hips, sex, average and range of age at diagnosis, follow-up
age, side of affected hip, type of treatment, and final
radiographic outcome.
Discrepancies of the extracted data and eligibility of
studies for inclusion were resolved by discussion with a
third reviewer (C.T.M.). Nine authors were contacted to
furnish additional information regarding their published
results. Three additional studies were included for anal-
ysis through this method.
Disease Severity Measures
Patients were grouped as having mild or severe
disease presentation. Under the Herring classification,
those individuals with class A or B radiographs repre-
sented mild progression of the disease, whereas those with
B/C or C were grouped in the more severe stage.14 For
studies using the Catterall classification, type 1 or 2 rep-
resented a mild presentation, whereas type 3 or 4 depicted
a more severe presentation.1
Outcome Measures
The primary data gathered from the studies con-
sisted of radiographic outcome defined by the Mose or
Stulberg classification. To pool the data for statistical
analyses, the radiographic outcomes were dichotomized
as good or poor. Under the Mose criteria, a good out-
come consists of a spherical femoral head that has 0 mm
in deviation of concentric circles between the frog-leg
lateral and anteroposterior views, whereas a fair outcome
r 2012 Lippincott Williams & Wilkins
J Pediatr Orthop ! Volume 32, Number 7, October/November 2012
has <2 mm in deviation. In contrast, a poor outcome
comprises a nonspherical femoral head with variability
>2 mm between the 2 projected radiographic views.15
A good radiographic outcome under a Stulberg I or II
classification is defined by a spherical femoral head that
has a normal hip joint or has a larger femoral head,
shorter femoral neck, or steep acetabulum (Table 1).8
Poor outcomes correlate with radiographic grades of
Stulberg III, IV, or V that demonstrate a nonspherical
femoral head in conjunction with abnormal features in
the femoral neck and acetabulum. Radiographs demon-
strating the different stages of Stulberg classification are
shown in Figure 1. Consistent with previous studies that
utilized the Stulberg or Mose classifications to assess ra-
diographic outcome, a Stulberg I or II grade is equivalent
to a good or fair Mose rating and was categorized as a
good outcome in our analysis, whereas Stulberg III-V was
equated to a poor Mose rating and categorized as a poor
outcome (Table 1).8,14,16–23
Statistical Methods
Meta-analysis was conducted using Comprehensive
Meta-Analysis v2.0 (Biostat, Englewood, NJ). Confidence
limits were set at 95% and data were pooled in a random
effect model. In addition, statistical analysis using a bi-
nary logistic regression model was performed with
PASW/SPSS v.18 (Predictive Analytic Software/Stat-
istical Package for the Social Sciences) (Chicago, IL) to
analyze the pooled patient-level data. The binary logistic
regression model was used with radiologic outcome
(good/poor), derived from the Stulberg and Mose classi-
fications, as the dichotomous outcome variable. Four
predictor variables were used: sex, age at treatment (di-
chotomized into less than 6 y or 6 y and older), initial
radiographic status defined by Herring and Caterall cri-
teria (dichotomized into mild or severe), and treatment
(dichotomized into operative or nonoperative). The use of
binary logistic regression afforded us the opportunity to
pool and use patient-level data from the studies in de-
termining predictors of the dichotomized radiographic
outcomes. Odds ratios (OR) along with 95% confidence
TABLE 1. Description of Stulberg and Mose Classification
Stulberg Class Features
I Normal hip joint
II Spherical femoral head with Z1 abnormalities:
coxa magna
shorten femoral neck
steep acetabulum
III Nonspherical but not flat femoral head with
Z1 abnormalities:
coxa magna
shorten femoral neck
steep acetabulum
IV Flat femoral head and abnormal femoral head,
neck, and acetabulum
V Flat femoral head with normal femoral neck and
acetabulum
Adapted from Stulberg et al.8
r 2012 Lippincott Williams & Wilkins
Operative Versus Nonoperative Treatments
intervals (CIs) were obtained. A sensitivity analysis was
performed on all eligible studies that used hip spica as the
sole nonoperative technique or Chiari pelvic osteotomy
(CPO) as the operative treatment modality. This was
carried out to assess whether or not hip spica casting
influenced the overall analysis and change any of the
possible predictor variables in measuring outcome varia-
bles as the method does not allow molding of the femoral
head within the acetabulum because of lack of weight
bearing compared with other nonoperative treatment
methods. Patients who are selected for CPO tended to
have a severe disease profile (Catterall 3 and 4), thus a
sensitivity analysis was performed to detect whether or
not exclusion of the treatment modality would alter any
of the possible predictor variables to be nonsignificant in
the overall analysis.
A 6-year age threshold was chosen as a cut-off for
the pediatric population, based on previous studies that
have demonstrated that children who present with LCPD
before the age of 6 fared better in radiographic outcomes
than those who acquire the disease at a later age.24–26
Operative treatments were categorized as: “femoral,”
“pelvic,” or a “combined procedure.” A “femoral” pro-
cedure consisted of femoral varus osteotomy (FVO), a
“pelvic” procedure of an innominate osteotomy (IO),
shelf acetabuloplasty/arthroplasty, or CPO, whereas a
“combined” procedure used both femoral and pelvic
procedures on the same affected hip.
To measure heterogeneity between the studies, Com-
prehensive Meta-Analysis was used to calculate the I2 sta-
tistic. An I2 value of 0%, 25%, 50%, and 75% represents 0,
low, moderate, and high heterogeneity, respectively.27
RESULTS
Literature Search
Our electronic search strategy resulted in 2069 ar-
ticles, of which 88 studies were deemed relevant for full-
text evaluation. Six additional studies were found from
textbook references. Out of the 94 studies, 23 met the
inclusion criteria (Fig. 2). One included study28 reported
Femoral Head Mose Prognosis
Spherical congruency Good Good
Spherical congruency Fair Good
Ovoid, mushroom shaped Poor Poor—slower onset of
osteoarthritis
Flat Poor Poor—moderate onset of
osteoarthritis
Flat Poor Poor—early onset of
osteoarthritis
www.pedorthopaedics.com | 699
Nguyen et al J Pediatr Orthop ! Volume 32, Number 7, October/November 2012

FIGURE 1. Frog-leg radiographs of Stulberg I to V. A and B represent Stulberg I and II (good results). C, D, and E represent
Stulberg III, IV, and V (poor results).

some of the same patients included in another long-term
paper10; data from these 2 studies were extracted to yield
the latest results without including the same patients in
the analysis twice.
Of the excluded studies, 28 did not report radio-
graphic outcome using Mose or Stulberg criteria, 24 did
not report age at diagnosis or at treatment, 8 studies in-
volved >1 treatment procedures where outcomes were
not differentiated by treatment, 5 studies described out-
comes with other hip conditions, 3 studies had <10 pa-
tients reported, and 3 did not use the Catterall or Herring
classification as an initial assessment.
Population Characteristics
Twelve of the 23 eligible studies evaluated radio-
graphic outcomes at skeletal maturity.3,4,10,16,19–21,29–33
Radiographic assessment using Mose or Stulberg criteria of
LCPD patients ranged from 1 to 64 years after treatment.

Citation
Identified
electronically
(n=2,069)

Article
considered for Citation
Full Text Identifed
Review manually
(n=94) (n=6)

Studies Excluded:
Did not use Mose/Stulberg (n=28)
Did not use Catterall/Herring (n=3)
Less than 10 patients (n=3)
Missing age of diagnosis/treatment (n=24)
Other concurrent hip disorders (n=5)
Treatment not specified (n=8)

Total Studies
included in
Meta-Analysis
(n=23)

FIGURE 2. Study selection process identifying eligible studies.

700 | www.pedorthopaedics.com r 2012 Lippincott Williams & Wilkins

J Pediatr Orthop ! Volume 32, Number 7, October/November 2012
Ten studies evaluated operative procedures,18,20,23,30,31,33–37
7 studies examined nonoperative treatment,4,10,22,24,28,29,32
and 6 studies compared operative and nonoperative treat-
ments.3,16,19,21,38,39 All studies included in the meta-analysis
were retrospective in design.
The sample size of eligible studies ranged from 10 to
358 patients, which yielded 1232 patients and 1266 hips
for analysis. A total of 783 hips were treated non-
operatively and 483 hips were treated operatively. At the
time of initial evaluation, patients ranged from 1.3
through 15.2 years old. Among studies that listed the side
of affected hip, the ratio distribution of left to right was
165:183. For studies that detailed patients’ sex, there were
599 (87%) male and 91 (13%) female patients. Table 2
describes the population characteristics of the included
studies.
Statistical Analysis
Meta-analysis based on study-level data examining
treatment type (operative or nonoperative) was performed
on the eligible 23 studies. This subgroup analysis of treat-
ment type compared radiographic outcome among 783 hips
treated with nonoperative methods to 483 hips that un-
derwent operative treatment. The success rate of treatment
for operative treatment was 51.7%, whereas nonoperative
treatment had a success rate of 52.2% in restoring femoral
head sphericity (Fig. 3). Without taking into account the
effects of any other variables, such as sex, age, or disease
severity, the analysis revealed no significant differences be-
tween the 2 treatment types (P = 0.678). The I2 values for
operative and nonoperative treatment were 75% and 91%,
respectively, thus indicating high heterogeneity across the
studies. With such high levels of variability across and
within the studies, further statistical analysis was performed
to incorporate the effects of multiple variables through
regression models.
The results from the logistic regression analyses in-
dicated that after controlling for age, sex, and disease
severity, patients who had operative treatment were about
twice as likely as those who had nonoperative treatment
to have a good outcome (OR = 2.01; P = 0.001; 95% CI,
1.309-3.09). Similarly, after adjusting for sex, disease se-
verity, and treatment type, patients of age less than 6
years were almost 4 times as likely as those 6 years or
older to have a good radiographic outcome (OR = 3.882;
P < 0.0001; 95% CI, 2.026-7.36).
Controlling for treatment type, patients less than 6
years old with mild disease severity were 15 times as likely
to have a good radiographic outcome (OR = 15.564;
P < 0.0001; 95% CI, 7.377-32.937) than their counterparts
with severe disease condition. When disease severity was
controlled for, patients younger than six years did not have
any difference in radiographic outcomes when operative
treatment was compared with nonoperative treatment
(OR = 1.071; P = 0.828; 95% CI, 0.582-1.968). Con-
versely, adjusting for disease severity among patients age 6
or older, operative treatment was almost twice as likely to
yield a good radiographic outcome in comparison with
nonoperative treatment (OR = 1.754; P < 0.0001; 95% CI,
r 2012 Lippincott Williams & Wilkins
Operative Versus Nonoperative Treatments
1.299-2.370). Patients with mild disease severity were twice
as likely as those with severe disease presentation to have a
good radiographic outcome than a poor outcome among
patients 6 years or older (OR = 2.014; P < 0.0001; 95%
CI, 1.427-3.842).
Furthermore, controlling for sex, treatment, and
age, patients with lesser rather than greater disease se-
verity were twice as likely to have a good rather than poor
outcome (OR = 2.036; P = 0.001; 95% CI, 1.326-3.132).
A patient having a good or bad outcome was not de-
pendent on or associated with sex. Males were equally as
likely as females to have good or bad outcomes regardless
of age, treatment, and initial disease severity (OR =
1.248; P = 0.486; 95% CI, 0.670-2.325).
Additional results from subgroup analysis using the
w2 test of association indicated a statistically significant
relationship between surgical types [femoral (FVO)/pelvic
(shelf, IO, or CPO) procedures] and the radiologic out-
comes for patients at age less than 6. For age at treatment
less than 6 years, a patient treated with pelvic rather than
a femoral procedure was about 5 times as likely to have
good radiographic outcomes (w2 = 4.488; P = 0.034; un-
adjusted OR = 5.20; 95% CI, 1.021-26.471). For patients
aged 6 or older, pelvic procedures were equally as likely as
femoral procedures to yield a successful radiographic
outcome (w2 = 1.845; P = 0.174; unadjusted OR = 1.329;
95% CI, 0.881-2.004).
A sensitivity analysis was performed on studies that
included hip spica casting or CPO as treatment modality
to assess whether or not the inclusion or exclusion of
these methods would affect the other variables in pre-
dicting radiographic outcomes. Using cases that excluded
treatment with a hip spica cast in the sensitivity analysis;
both Herring (P = 0.008) and Caterall (P = 0.027) clas-
sifications significantly predicted radiographic outcomes
regardless of treatment. Thus, the inclusion or exclusion
of hip spica cases did not alter the result where initial
severity predictors (Herring/Catterall) were still sig-
nificant as measures of disease severity in predicting
outcome. Similarly, excluding CPO cases, initial severity
measures [Herring (P = 0.002) and Caterall (P = 0.025)]
were significant predictors of outcome. This suggested
that our results were robust with or without CPO as
treatment modality when Herring or Caterrall was used
as initial disease severity measures.
DISCUSSION
Children diagnosed with LCPD have an increased
risk in developing debilitating osteoarthritis by early
adulthood. To prevent or delay the onset of osteo-
arthritis, the goal of nonoperative or operative treatment
has focused on restoring a spherical femoral head.
However, despite the plethora of literature on non-
operative and operative methods in treating LCPD, there
still exists a lack of consensus among studies regarding
the risk factors that influence patient outcomes and the
optimal treatment of the condition. Without taking into
account any risk factors, our pooled meta-analysis of
www.pedorthopaedics.com | 701
Nguyen et al J Pediatr Orthop ! Volume 32, Number 7, October/November 2012

TABLE 2. Study Population Characteristics

Range Mean
Onset Age Length Range
Mean Age Range of Mean Age Treated Follow-up Follow-up No. No. Good Female
References Intervention Onset (y) Age (y) Treated (y) (y) (y) (y) Patients Hips Outcome (%)
Alves and Nonoperative 6.8 4-11.0 — — 19 10-33.0 43 49 27 19
Santili35
Arkader et al16 Nonoperative, 10.8 9-14.0 — — 10 5-22.0 43 43 7 9.30
FVO, CPO
Crutcher and FVO 7.6 4.0-10.5 8.33 4.5-10.8 8 5-10.5 14 14 7 14.29
Staheli18
Ghanem et al36 Shelf, FVO 7.5 6.2-11.0 8.6 7.1-13.0 9.5 5.2-12 30 30 19 26.67
Gower and Nonoperative 8.3 2-15.0 8.9 2.9-14 44.6 30-48 36 37 — 13.89
Johnston28w
Ippolito et al29 Nonoperative 7.5 2.4-15.0 — — 25 7-56 61 61 25 11.48
Kamegaya FVO, 9.2 6.25-11.25 — — — 16-20* 36 36 17 0.00
et al19 nonoperative
Kamegaya FVO 8.4 4.1-12.1 8.7 4.4-12.2 18.5* SM 37 38 20 8.11
et al30
Kelly et al4 Nonoperative 6.3 2-12.0 — — 22.4 SM 80 80 73 —
Kim et al22 Nonoperative 7.1 3.5-11.1 — — 7.8 6.3-10.8 20 20 17 15.00
Kruse et al38 Nonoperative 7 4.0-10.0 — — 28 7-45 17 18 11 13.89
Shelf 8 2.0-14.0 11 7-15.0 19 2-47 19 20 14
McAndrew and Nonoperative 8.2 4.7-15.0 — — 47.7 39-64 28 28 14 14.29
Weinstein10w
Noonan et al20 FVO — — 10.6 9.5-13.9 10 4.2-17.8 17 18 6 29.41
Oh et al37 Shelf — — 8.7 6-12.0 5 3-14.0 20 20 8 20.00
Osman et al3 Nonoperative, 9.75 8-14.3 — — 6.4 2.2-20 44 48 9 18.18
FVO, shelf
Pecquery et al34 Shelf 7.2 2.9-12.4 8 4.5-13.0 4.25 1.0-8.75 21 21 14 9.52
Reddy and CPO — — 8.5 6.0-11.0 6.1 — 21 22 15 19.05
Morin31
Robinson et al32 IO 6.25 2.6-12.2 8 4.8-13.2 8.2 5.0-16.3 25 25 23 7.41
Rosenfeld et al24 Nonoperative 4.6 2-5.9 — — — > 1.0 172 188 152 —
Segev et al33 Arthrodiatasis — — 12.3 9.4-15.1 5.7 4.3-7.8 10 10 0 23.08
Sharma et al21 Nonoperative, 6.75 4-8.0 — — — SM 28 28 7 14.29
FVO, shelf,
IO
Sponseller et al23 FVO, IO 7.2 3.5-13.2 — — 8.9 5-13 88 91 62 —
Wiig et al39 Nonoperative, 5.8 1.3-15.2 — — — > 5.0 358 358 176 —
FVO
*Mean age range at follow-up.
wGower and McAndrew studies had overlapping patients.
CPO indicates Chiari pelvic osteotomy; FVO, femoral varus osteotomy; IO, innominate osteotomy; SM, skeletal maturity.

1266 hips found that both operative and nonoperative
procedures had the same likelihood of obtaining a good
result. However, because of the high level of heterogeneity
across the studies and the need to assess the influence of
multiple risk factors that may influence the outcome of a
patient with LCPD, a binary logistic regression model
was performed to examine which modality of treatment
yielded a more superior result among nonoperative and
operative procedures. Nonoperative treatment of LCPD
has a tractable history and long tradition. The pioneer
physicians—Legg, Calvé, and Perthes, all treated their
patients nonoperatively. These treatment options in-
cluded: no treatment except symptom management, bed
rest, range of motion therapy, weight relief (splints,
traction, crutches), or containment of the femoral had
using abduction splint or Petrie casting.40 This meta-
analysis found that patients who are younger than 6 years
had the same likelihood in obtaining a good radio-
graphic outcome with either nonoperative or operative
methods. In contrast, patients age 6 or older who were
treated nonoperatively fared worse than those treated
with surgery. This finding is consistent with previous
studies that suggested that nonoperative treatment yields
favorable outcome among younger patients with mild
presentation of the disease, but ineffective especially
among older patients with known risk factors for poor
prognosis.24,41–44
Studies that have examined outcomes after surgical
treatments, such as FVO, IO, shelf acetabuloplasty/
arthroplasty, and Chiari pelvic osteomy have found better
results for their patients in comparison with those who
were treated nonoperatively among both the age
groups.3,16,21,39,45 Our meta-analysis revealed an associa-
tion between radiographic outcomes and the type of
procedure performed, especially among patients with
early-onset LCPD. In a subgroup analysis with regard to
operative procedures and age, we found that patients
younger than 6 years old, treatment with a pelvic procedure

702 | www.pedorthopaedics.com r 2012 Lippincott Williams & Wilkins

J Pediatr Orthop ! Volume 32, Number 7, October/November 2012 Operative Versus Nonoperative Treatments

Study Subgroup within study Event Event rate and 95% CI

rate
Successful Outcomes /
Total Subjects

Alves Non-Operative 27 / 49 0.55

Arkader Non-Operative 3 / 21 0.14
Ippolito Non-Operative 25 / 61 0.41
Kamegaya Non-Operative 4 / 18 0.22
Kelly Non-Operative 73 / 80 0.91
Kim Non-Operative 17 / 20 0.85
Kruse Non-Operative 13 / 18 0.72
McAndrew Non-Operative 14 / 28 0.50
Osman Non-Operative 4 / 26 0.15
Rosenfeld Non-Operative 152 / 188 0.81
Sharma Non-Operative 3 / 14 0.21
Wiig Non-Operative 134 / 260 0.52
Pooled Estimate of Non-Operative Treatment 469 / 783 0.52
Arkader Operative 4 / 22 0.18
Kamegaya Operative 20 / 38 0.53
Kruse Operative 14 / 20 0.70
Osman Operative 5 / 22 0.23
Sharma Operative 4 / 14 0.29
Wiig Operative 42 / 98 0.43
Crutcher Operative 7 / 14 0.50
Ghanem Operative 19 / 30 0.63
Kamegaya Operative 13 / 18 0.72
Noonan Operative 6 / 18 0.33
Oh Operative 8 / 20 0.40
Pecquery Operative 14 / 21 0.67
Reddy Operative 15 / 22 0.68
Robinson Operative 23 / 25 0.92
Segev Operative 0 / 10 0.05
Sponseller Operative 62 / 91 0.68
Pooled Estimate of Operative Treatment 256 / 483 0.52
0.00 0.50 1.00

FIGURE 3. Effect of operative and nonoperative treatment based on radiographic outcomes. Without taking into consideration
any other modulating variables, the subgroup indicated that operative and nonoperative treatment yielded successful radiograph
outcomes at the rate of 51.7% and 52.2%, respectively. A high level of heterogeneity was found at the study-level data analysis.
CI indicates confidence interval.

was 5 times as likely to result in a spherical femoral head as
treatment with a femoral procedure. Pelvic procedures ap-
pear to provide better shelf coverage and containment than
femoral techniques, thus allowing the femoral head to be
molded within the acetabulum.16 However, among patients
6 years or older, pelvic procedures had the same likeli-
hood of yielding a good radiographic outcome as femoral
procedures.
This meta-analysis strongly suggests that patients
who present with LCPD at an early age younger than 6,
tended to have a more favorable outcome than those di-
agnosed at a later age. Previous studies have suggested
that this may be because of greater time available for the
femoral head to remodel before skeletal maturity is
reached.10,21,46 We also found that the initial stage of
disease presentation is another significant predictor of
long-term radiographic outcome. Patients with a milder
form of LCPD at the time of diagnosis and treatment
have significantly better radiographic outcomes that those
with a more severe involvement. Femoral heads that were
not as deformed at initial presentation tended to retain
their spherical shape and remain congruent within the
acetabulum.
Although males are more prone to develop LCPD,
the influence of sex on outcome has been a topic of dis-
agreement in the literature. Some studies suggest a female
sex predisposed the patient to a worse prognosis,2–5
whereas others found that sex did not influence out-
comes.24,39,47 In this meta-analysis, males outnumbered
female patients at a rate of 6.5 to 1. However, when the
influence of age, disease severity, and treatment methods
were controlled in our analysis, sex was not a significant
predictor of long-term radiographic outcome.
This meta-analysis focused on radiographic out-
comes. The studies that met our inclusion criteria did not
report clinical outcomes in a standardized manner that
would enable data pooling. With regard to radiographic
outcome, femoral heads that are more nonspherical and
incongruent at skeletal maturity are at a higher risk for
developing debilitating osteoarthritis.8,48–50 In a long-
term study, Stulberg and colleagues reported a correlation
between clinical and radiographic outcome among LCPD

r 2012 Lippincott Williams & Wilkins www.pedorthopaedics.com | 703

Nguyen et al J Pediatr Orthop
patients. Stulberg et al8 found that hips categorized as
Stulberg I and II developed osteoarthritis at a low rate
ranging between 0% to 7%, whereas those belonging to
Stulberg III, IV, and V have a higher rate of degenerative
joint disease with 36%, 67%, and 81%, respectively.
These evaluations at the end of bone remodeling suggest a
good correlation between radiographic outcome and the
clinical risk associated with degenerative joint disease.
We advocate for a more standardized system to
grade clinical outcomes and functional mobility in the
long-term follow-up of LCPD patients. Until such a
system has been developed, the evaluation of treatment
efficacy will continue to focus on the systematic grading
of radiographs.
Study Strengths and Limitations
This meta-analysis examined 23 studies and 1266
hips that met the inclusion criteria. Follow-up ranged
from 1 to 64 years after treatment with studies published
between 1971 and 2010. Ideally, the final radiographic
follow-up should be performed at skeletal maturity to
better predict the risk of degenerative changes. Although
more than half of the included studies evaluated patients
at skeletal maturity, a few of the included studies assessed
treatment outcome before epiphyseal plate closure. On a
related note, there is a need for more longitudinal studies
that assess patients beyond just skeletal maturation but
also into their late 40s and 50s to assess the long-term
outcomes of various treatments.
All 23 of the included studies were retrospective in
nature. Although this meta-analysis is subject to the same
limitations or potential for bias as the pooled studies, the
performance of a binary logistic regression analysis em-
powered us to derive important conclusions from the
pooled data.
We only used studies that included patient-level
data and those with age ranges that can be grouped as
early disease onset (younger than 6) or as late onset (age 6
and older) to pool and analyze the data in a statistically
valid manner. This is the first meta-analysis to utilize lo-
gistic regression to assess LCPD treatment. One of the
advantages of employing logistic regression in analyzing
pooled patient-level data is the ability to achieve a more
stable and robust result. The statistical use of logistic
regression allows us to examine multiple predictor vari-
ables that may influence the long-term outcomes of
LCPD patients and adjust for their individual effects.
CONCLUSIONS
Among patients younger than 6 years, operative
procedures had the same likelihood of a good outcome as
nonoperative treatment. Comparing operative treatment
among patients younger than 6 years, pelvic procedures
yielded significantly better radiographic results when
compared with femoral procedures. However, for patients
6 years or older, operative treatment was almost twice as
likely to result in a successful radiographic outcome as
nonoperative treatment. The likelihood of a successful
radiographic outcome among patients 6 years or older
704 | www.pedorthopaedics.com
! Volume 32, Number 7, October/November 2012
was equivalent for femoral and pelvic procedures. We
strongly advocate early diagnosis and treatment of pa-
tients who present with LCPD as delay in treatment may
decrease the likelihood of a favorable outcome.
ACKNOWLEDGMENTS
The authors thank Tony Herring, MD, Alex Arkader,
MD, and Makoto Kamegaya, MD for generously providing
additional data on published studies. The authors also thank
Debra Orr, MSS, MLIS and Douglas Mann, PhD for
assistance in preparing the manuscript.
REFERENCES
1. Catterall A. The natural history of Perthes’ disease. J Bone Joint
Surg Br. 1971;53:37–53.
2. Daly K, Bruce C, Catterall A. Lateral shelf acetabuloplasty in
Perthes’ disease. A review of the end of growth. J Bone Joint Surg
Br. 1999;81:380–384.
3. Osman MK, Martin DJ, Sherlock DA. Outcome of late-onset
Perthes’ disease using four different treatment modalities. J Child
Orthop. 2009;3:235–242.
4. Kelly FB Jr, Canale ST, Jones RR. Legg-Calvé-Perthes disease.
Long-term evaluation of non-containment treatment. J Bone Joint
Surg Am. 1980;62:400–407.
5. Tempelton J, Green AD. Perthes’ disease—a long term follow-up.
Ulster Med J. 1983;52:125–130.
6. Rowe SM, Jung ST, Lee KB, et al. The incidence of Perthes’ disease
in Korea. A focus on differences among races. J Bone Joint Surg Br.
2005;87:1666–1668.
7. Nelitz M, Lippacher S, Krauspe R, et al. Perthes disease: current
principles of diagnosis and treatment. Dtsch Arztebl Int.
2009;106:517–523.
8. Stulberg SD, Cooperman DR, Wallensten R. The natural history of
Legg-Calvé-Perthes disease. J Bone Joint Surg Am. 1981;63:1095–1108.
9. Yrjonen T. Prognosis in Perthes’ disease after noncontainment
treatment. 106 hips followed for 28–47 years. Acta Orthop Scand.
1992;63:523–526.
10. McAndrew MP, Weinstein SL. A long-term follow-up of Legg-
Calvé-Perthes disease. J Bone Joint Surg Am. 1984;66:860–869.
11. Forster MC, Kumar S, Rajan RA, et al. Head-at-risk signs in Legg-
Calvé-Perthes disease: poor inter- and intra-observer reliability. Acta
Orthop. 2006;77:413–417.
12. Sambandam SN, Gul A, Shankar R, et al. Reliability of radiological
classifications used in Legg-Calvé-Perthes disease. J Pediatr Orthop
B. 2006;15:267–270.
13. Agus H, Kalenderer O, Eryanlmaz G, et al. Intraobserver and
interobserver reliability of Catterall, Herring, Salter-Thompson and
Stulberg classification systems in Perthes disease. J Pediatr Orthop
B. 2004;13:166–169.
14. Herring JA, Kim HT, Browne R. Legg-Calvé-Perthes disease. Part I:
classification of radiographs with use of modified lateral Pillar and
Stulberg classifications. J Bone Joint Surg Am. 2004;86:2103–2120.
15. Mose K. Legg-Calvé-Perthes Disease: A comparison between three
methods of conservative treatment. Universitetsforlaget I Aarhus.
1964. Thesis.
16. Arkader A, Sankar WN, Amorim RM. Conservative versus surgical
treatment of late-onset Legg-Calvé-Perthes disease: a radiographic
comparison at skeletal maturity. J Child Orthop. 2009;3:21–25.
17. Freeman RT, Wainwright AM, Theologis TN, et al. The outcome of
patients with hinge abduction in severe Perthes disease treated by
shelf acetabuloplasty. J Pediatr Orthop. 2008;28:619–625.
18. Crutcher JP, Staheli LT. Combined osteotomy as a salvage
procedure for severe Legg-Calvé-Perthes disease. J Pediatr Orthop.
1992;12:151–156.
19. Kamegaya M, Saisu T, Ochiai N, et al. A paired study of Perthes’
disease comparing conservative and surgical treatment. J Bone Joint
Surg Br. 2004;86:1176–1181.
r 2012 Lippincott Williams & Wilkins
J Pediatr Orthop ! Volume 32, Number 7, October/November 2012
20. Noonan KJ, Price CT, Kupiszewski SJ, et al. Results of femoral
varus osteotomy in children older than 9 years of age with Perthes
disease. J Pediatr Orthop. 2001;21:198–204.
21. Sharma S, Shewale S, Sibinski M, et al. Legg-Calvé-Perthes disease
affecting children less than eight years of age: a paired
outcome study. Int Orthop. 2009;33:231–235.
22. Kim WC, Hosokawa M, Tsuchida Y, et al. Outcomes of new pogo-
stick brace for Legg- Calvé-Perthes disease. J Pediatr Orthop B.
2006;15:98–103.
23. Sponseller PD, Desai SS, Millis MB. Comparison of femoral and
innominate osteotomies for the treatment of Legg-Calvé-Perthes
disease. J Bone Joint Surg Am. 1988;70:1131–1139.
24. Rosenfeld SB, Herring JA, Chao JC. Legg-Calvé-Perthes disease: a
review of cases with onset before six years of age. J Bone Joint Surg
Am. 2007;89:2712–2722.
25. Canavese F, Dimeglio A. Perthes’ disease: prognosis in children
under six years of age. J Bone Joint Surg Br. 2008;90:940–945.
26. Jani LF, Dick W. Results of three different therapeutic groups in
Perthes’ disease. Clin Orthop Relat Res. 1980;150:88–94.
27. Higgins JP, Thompson SG, Deeks JJ, et al. Measuring inconsistency
in meta-analyses. Br Med J. 2003;327:557–560.
28. Gower WE, Johnston RC. Legg-Perthes disease. Long-term follow-
up of thirty-six patients. J Bone Joint Surg Am. 1971;53:759–768.
29. Ippolito E, Tudisco C, Farsetti P. The long-term prognosis of
unilateral Perthes’ disease. J Bone Joint Surg Br. 1987;69:243–250.
30. Kamegaya M, Saisu T, Takazawa M, et al. Arthrographic indicators
for decision making about femoral varus osteotomy in Legg-Calvé-
Perthes disease. J Child Orthop. 2008;2:261–267.
31. Reddy RR, Morin C. Chiari osteotomy in Legg-Calvé-Perthes
disease. J Pediatr Orthop B. 2005;14:1–9.
32. Robinson HJ Jr, Putter H, Sigmond MB, et al. Innominate
osteotomy in Perthes disease. J Pediatr Orthop. 1988;8:426–435.
33. Segev E, Ezra E, Wientroub S, et al. Treatment of severe late-onset
Perthes’ disease with soft tissue release and articulated hip distraction:
revisited at skeletal maturity. J Child Orthop. 2007;1:229–235.
34. Pecquery R, Laville JM, Salmeron F. Legg-Calvé-Perthes disease
treatment by augmentation acetabuloplasty. Orthop Traumatol Surg
Res. 2010;96:166–174.
35. Alves MW, Santili C. Long-term analysis of a conservative
approach in Legg-Calvé-Perthes disease. Acta Ortop Bras.
2005;13:238–244.
36. Ghanem I, Haddad E, Haidar R, et al. Lateral shelf acetabuloplasty
in the treatment of Legg-Calvé-Perthes disease: improving mid-term
outcome in severely deformed hips. J Child Orthop. 2010;4:13–20.
r 2012 Lippincott Williams & Wilkins
Operative Versus Nonoperative Treatments
37. Oh CW, Rodriguez A, Guille JT, et al. Labral support shelf
arthroplasty for the early stages of severe Legg-Calvé-Perthes
disease. Am J Orthop. 2010;39:26–29.
38. Kruse RW, Guille JT, Bowen JR. Shelf arthroplasty in patients who
have Legg-Calvé-Perthes disease. A study of long-term results.
J Bone Joint Surg Am. 1991;73:1338–1347.
39. Wiig O, Terjesen T, Svenningsen S. Prognostic factors and out-
come of treatment in Perthes’ disease: a prospective study of 368
patients with five-year follow-up. J Bone Joint Surg Br. 2008;90:
1364–1371.
40. Hefti F, Clarke NM. The management of Legg-Calvé-Perthes’
disease: is there a consensus? A study of clinical practice preferred by
the members of the European Paediatric Orthopaedic Society.
J Child Orthop. 2007;1:19–25.
41. Carney BT, Minter CL. Nonsurgical treatment to regain hip
abduction motion in Perthes disease: a retrospective review. South
Med J. 2004;97:485–488.
42. Brotherton BJ, McKibbin B. Perthes’ disease treated by prolonged
recumbency and femoral head containment: a long-term appraisal.
J Bone Joint Surg Br. 1977;59:8–14.
43. Meehan PL, Angel D, Nelson JM. The Scottish Rite Abduction
Orthosis for the treatment of Legg-Perthes Disease. J Bone Joint
Surg. 1992;74:2–12.
44. Martinez AG, Weinstein SL, Dietz FR. The weight bearing
abduction brace for the treatment of Legg Perthes Disease. J Bone
Joint Surg. 1992;74:12–21.
45. Poussa M, Yrjonen T, Hoikka V, et al. Prognosis after conservative
and operative treatment in Perthes’ disease. Clin Orthop Relat Res.
1993;297:82–86.
46. Engelhardt P. Late prognosis in Perthes’ disease: which factors
determine arthritis risk? Z Orthop Ihre Grenzgeb. 1985;123:168–181.
47. Guille JT, Lipton GE, Szoke G, et al. Legg-Calvé-Perthes Disease in
girls. A comparison of the results with those seen in boys. J Bone
Joint Surg Am. 1998;80:1256–1263.
48. Lecuire F. Long-term outcome of primary osteochondritis of the hip
(Legg-Calvé-Perthes’ disease). J Bone Joint Surg Br. 2002;84:
636–640.
49. Mose K. Methods of measuring in Legg-Calvé-Perthes disease with
special regard to the prognosis. Clin Orthop Relat Res.
1980;150:103–109.
50. Salter RB, Thompson GH. Legg-Calvé-Perthes disease. The
prognostic significance of the subchondral fracture and a two-group
classification of the femoral head involvement. J Bone Joint Surg
Am. 1984;66:479–489.
www.pedorthopaedics.com | 705