This is a prospective study performed over a five-year period approved by the local institutional review board. All subjects and their parents were informed about the study and provided written informed consent prior to enrollment.
Patients
New and existing patients with Legg-Calvé-Perthes disease treated at the Division of Pediatric Orthopaedics of our institution from April 2006 through April 2011 were enrolled. Inclusion criteria were (1) Legg-Calvé-Perthes disease after complete reossification of the epiphysis, and (2) MRI evaluation of the affected hip. Exclusion criteria were incomplete reossification of the epiphysis on radiographs (unhealed heads), inadequate MRI protocol, stainless steel implants too close to the hip that could interfere with MRI quality, or previous hip surgery that included opening of the capsule.
During the study period, 134 patients with Legg-Calvé-Perthes disease who had complete radiographic documentation were seen at our institution. All had pelvic radiographs made in the anteroposterior view and frog-leg lateral (Lauenstein) view30,31. An elongated neck lateral radiograph (90° of flexion, 20° to 45° of abduction, and neutral rotation—similar to the Dunn-Imhäuser view)32,33 was made for thirty-six patients. Thirty-five patients were excluded because of incomplete epiphyseal healing. Therefore, a radiographic series of ninety-nine patients diagnosed with healed Legg-Calvé-Perthes disease was selected. MRI was not performed in forty-one patients because of a failed contact or irregular follow-up (thirty patients) or because stainless steel implants, such as plates and screws, bordered the hip joint (eleven patients). MRI was performed despite the presence of Kirschner wires at the iliac osteotomy site.
Fifty-eight subjects underwent MRI evaluation of the affected hip, but four were excluded because of inadequate MRI protocol or high-grade distortion secondary to metallic implants. After proper selection, a total of fifty-four patients with fifty-nine hips were evaluated (five patients had bilateral involvement). In eight patients, the cartilage evaluation was compromised because of minor metallic artifacts, but the labral evaluation was possible. Therefore, the acetabular labrum was evaluated in fifty-nine hips and the articular cartilage was evaluated in fifty-one hips (Fig. 1). The presence or absence of pain or stiffness in the hip, treatment method, or any other clinical issue did not influence inclusion or exclusion. For the patients previously treated with pelvic or femoral surgery, a minimum period of one year after surgery was required before an MRI examination was performed. For all analyses, it did not matter if morphologic changes in the femur and acetabulum were due to prior osteotomies or the underlying disease.
Thirty-one affected hips (53%) were on the right side, and twenty-eight (47%) were on the left side, but five patients had bilateral involvement. The mean age (and standard deviation) at the time of the MRI was 14.0 ± 2.8 years (median, fourteen years; range, eight to twenty-one years). The mean period from the clinical onset of Legg-Calvé-Perthes disease to the MRI was 8.0 ± 2.7 years (median, seven years; range, four to fifteen years). Nineteen (32%) of the fifty-nine hips had undergone conservative treatment (symptomatic relief, weight-bearing activity restriction, and physical therapy with hip adductor stretching). Forty hips (68%) underwent surgical treatment, and the procedures included Salter osteotomy (nineteen hips; 32%), femoral varus osteotomy (seven hips; 12%), arthrodiastasis (five hips; 9%), triple pelvic osteotomy (two hips; 3%), and Pemberton osteotomy (one hip; 2%). The remaining six hips (10%) underwent revision surgery because of a poor outcome after one of the above-mentioned procedures.
Radiographic Interpretation
On the anteroposterior pelvic radiograph, the affected hip was classified, according to the system of Stulberg et al.29, on the basis of hip morphology and congruence. The Stulberg class-I group was without femoral deformities. Coxa brevis was considered present when the length of the involved femoral neck (the distance from the center of the femoral head to the femoral neck base; Fig. 2-A) was at least 85% less than the normal side; coxa magna, when the involved femoral head had a diameter (Fig. 2-A) at least 10% greater than the normal side29; and coxa vara, when the femoral neck-shaft angle34 (Fig. 2-C) was <120°. The height of the greater trochanter29 (Fig. 2-D) was assessed to evaluate the possibility of extra-articular impingement. The extrusion index23 (Fig. 2-B) was measured and considered abnormal when it was >25%. Acetabular overcoverage was present when the lateral center-edge angle of Wiberg (Fig. 2-C) was >35°, and acetabular coverage deficiency was evident when the angle was <20°23. The radiographic alpha angle23 was measured in the lateral radiograph.
A subjective diagnosis of coxa magna or brevis was used for the five patients with bilateral involvement. The above-mentioned radiographic parameters were analyzed by one pediatric orthopaedic surgeon (nonblinded).
MRI Acquisition Protocol
Noncontrast MRI was performed with use of a 1.5-T system (Achieva; Philips Healthcare, Best, the Netherlands). Patients were positioned supine with the hips in neutral position and the feet aligned parallel and kept together with a position holder. Multiplanar evaluation was performed with two sequences of the pelvis with use of a body coil and four sequences of each hip with use of a small flexible surface coil to provide a small field and high spatial resolution.
The pelvic panoramic sequences were made in coronal and axial planes. The T1-weighted and T2-weighted sequences of the hip with fat suppression were made in the coronal (voxel: 0.7 × 0.9 × 3 mm), sagittal (voxel: 0.9 × 1.2 × 3 mm), and axial oblique (voxel: 0.8 × 1.2 × 3 mm) planes. The axial oblique scan was acquired parallel to the axis of the femoral neck.
MRI Interpretation
For the labrum, a classification system modified from Czerny et al.35 was used (see Appendix). The location and extent of abnormalities (see Appendix) were reported according to the clock-face nomenclature36,37. The anteromedial site of the acetabular rim was designated as three o’clock (3h); the superior site, as 12h; and the posterolateral site, as 9h. The regular clock-face nomenclature was used for the right hip, but a reverse clock-face projection was used for the left side so that the same nomenclature was used for both sides.
The articular cartilage of the acetabulum was considered normal if images showed a regular and homogeneous appearance. Cartilage delamination was considered to be present when one of two criteria were met: (1) a high-signal-intensity line between the hyaline cartilage and the subchondral bone was seen at the peripheral portion of the acetabulum in T2-weighted sequences with fat suppression, or (2) a peripheral and abnormal region of the hyaline cartilage showed focal linear low-signal intensity that was continuous with the surface of the central cartilage whose signal was normal on T1-weighted or T2-weighted fat-suppression sequences27,38-40. Focal substitution of the regular articular cartilage signal intensity by the high-signal intensity of the synovial fluid was considered a chondral defect24. Diffuse narrowing of the articular cartilage, with the joint space clearly reduced and a bright fluid signal, was classified as osteoarthritis.
All MRI studies were analyzed independently by one experienced musculoskeletal MRI radiologist, blinded to radiographic data and to the second reader, and one pediatric orthopaedic surgeon with previous training and clinical experience in the interpretation of MRI studies of the hip who was blinded to the first reader.
Labral and cartilage abnormalities were considered to occur only when they were seen in at least two consecutive slices in the same plane or at the same location in two different planes. MRI readers were aware of the possibility of a sublabral recess41,42 (see Appendix).
The alpha angle was measured in the axial oblique images passing through the center of the femoral head43 at position 3h. A best-fit circle was drawn over the cortical bone of the femoral head, and the point where the femoral head or neck extended beyond the circle was determined. The angle was measured using the femoral neck axis and a line connecting the center of the femoral head to the point of beginning asphericity of the head-neck contour22,43,44 (Fig. 3). Every measurement was performed three times by the same reader. When the femoral head was mostly aspherical, the alpha angle was measured using the most approximated circle based on the femoral head and acetabular morphology, as exemplified in Figure 4.
The acetabular version was assessed in the axial plane of pelvic panoramic sequences. The measurement was made with compensation of the pelvic obliquity and rotation, according to previous studies15,45, at the cranial part of the acetabulum over the position 1h to 11h or 2h to 10h.
Statistical Methods
Initially, the sample was divided into groups according to the Stulberg classification, and the distributions of frequency for each radiographic and MRI parameter were analyzed with the chi-square exact test.
The comparative group was represented by Stulberg class-I hips with normal morphology (an alpha angle of <55°; absence of coxa vara, magna, and brevis; and normal trochanteric height), without acetabular retroversion (five hips).
The MRI and radiographic alpha angles were compared for reliability analysis by a Bland-Altman approach46. The receiver operating characteristic curve method was applied to report sensitivity and specificity of the MRI alpha angle in relation to the MRI labral abnormalities with the purpose of estimating a cutoff value.
Intraobserver and interobserver agreement for labral and chondral MRI abnormalities was assessed through the weighted kappa coefficient47.
The association between labral and/or chondral MRI abnormalities and hip deformities (reduced offset; coxa vara, brevis, and magna; trochanteric height; cephalic extrusion; acetabular coverage; and version abnormalities) was analyzed using the chi-square or the chi-square exact test. The relative risk was estimated using odds ratio (OR) analysis and was reported if it increased. The same analyses were performed for hips with a normal alpha angle. With use of logistic regression for multivariate analysis, the adjusted OR was calculated for deformities significantly associated with MRI abnormalities of the acetabular labrum and cartilage with the purpose of achieving the relative risk when deformities coexisted.
For all statistical analyses, p values of <0.05 were considered significant.
Source of Funding
There was no external funding or financial support for this research. The MRI device was previously obtained with partial support from the governmental institution Research and Project Financing (FINEP).
For the fifty-four patients evaluated with MRI of the hip, the summary of the radiographic analyses is reported in Table I and the distribution among Stulberg groups is shown in Figure 5.
Acetabular retroversion was found in thirty-one (53%) of fifty-nine hips (five [50%] of ten Stulberg class-I hips; nine [53%] of seventeen class-II hips; nine [56%] of sixteen class-III hips; three [33%] of nine class-IV hips; and five [71%] of seven class-V hips). Acetabular retroversion was not found in the seven hips treated by varus osteotomy, but was found in 32% (six) of nineteen hips treated conservatively, 40% (two) of five hips treated by arthrodiastasis, and 90% (nineteen) of twenty-one hips treated by Salter or triple osteotomy.
Abnormalities of the acetabular labrum and cartilage were found on MRI scans of 75% (forty-four) of fifty-nine hips and 47% (twenty-four) of fifty-one hips, respectively. The overall frequency of labral and cartilage abnormalities identified on MRI scans was 75% (forty-four hips) and 45% (twenty-three hips), respectively, for reader 1 and 76% (forty-five hips) and 47% (twenty-four hips) for reader 2. Interobserver and intraobserver agreement was considered at least substantial for labrum and cartilage evaluation (interobserver weighted kappa, 0.82 and 0.81, respectively; intraobserver weighted kappa varied from 0.78 to 0.83)46.
The frequency of labral abnormalities included type I in two hips (3%), type II in twelve (20%), type III in nineteen (32%), and type IV in eleven (19%). Of the forty-four labral abnormalities, most were smaller than one quadrant (type A in four hips [9%] and type B in nineteen hips [43%]). The remaining hips had abnormalities extending greater than one quadrant (type C in twelve hips [27%] and type D in nine hips [20%]). Articular cartilage delamination, which was detected in fifteen (29%) of fifty-one hips, was the most frequent MRI abnormality of the acetabular cartilage. Most labral and acetabular cartilage abnormalities were located at the anterolateral aspect of the acetabulum (3h to 12h; Fig. 6). There was strong correlation between labral and acetabular cartilage abnormalities (p = 0.002; OR = 16). The MRI abnormalities of the acetabular labrum and cartilage did not change significantly between different durations of disease (four to fifteen years; t test, p = 0.21).
Estimating a normal range of alpha angle in patients with Legg-Calvé-Perthes disease, the receiver operating characteristic curve analyses showed a graphic of true-positive rates compared with false-positive rates, with an area under the curve of 0.74. The cutoff of 55° represented the best correlated value (sensitivity of 71% and specificity of 93%). The mean MRI alpha angle value was 71° ± 33° (range, 30° to 130°; median, 78°; 95% confidence interval [CI], 63° to 80°) for all samples, 45° ± 20° (range, 30° to 113°; median, 42°; 95% CI, 33° to 56°) for hips with a normal labrum, and 80° ± 32° (range, 31° to 130°; median, 92°; 95% CI, 70° to 90°) for hips with a labral abnormality on MRI. There was significant correlation between increased alpha angle and the presence of acetabular labral (p < 0.001) and articular cartilage (p < 0.01) MRI abnormalities (see Appendix).
Absolute values of the radiographic and MRI alpha angle showed no significant agreement (paired t test, p < 0.05; mean of differences, 5°; 95% CI, 0.5° to 10°), but there was effective correlation of variability (coefficient, 0.85). The alpha angle was not accurately measurable in oblique axial MRI plane or lateral hip radiographs in 32% (nineteen) of the fifty-nine hips because of a major loss of sphericity. This feature was not observed in Stulberg class-I or II hips, but it was present in 31% of Stulberg class-III hips and 85% to 89% of Stulberg class-IV or V hips.
The frequency distribution of MRI findings, according to the Stulberg classification and the presence or absence of acetabular retroversion, is shown in a table in the Appendix. A normal labrum is significantly correlated with the Stulberg class-I group without acetabular retroversion (p < 0.001). Labral abnormalities were associated with Stulberg class I with acetabular retroversion and with Stulberg class-II, III, IV, and V groups. Among these groups with hip deformities, no significant differences were detected for labral (p > 0.05) or cartilage abnormalities (p = 0.15).
When all radiographic deformities were considered independently (Table II), the increased alpha angle was the factor most significantly associated with both acetabular labral and cartilage abnormalities, followed by coxa brevis. Coxa magna and a higher greater trochanter also showed significant association with labral abnormalities only. However, as deformities coexist and are considered dependent variables, a logistic regression for multivariate analysis was used. In the presence of other deformities, the alpha angle showed the strongest correlation and the highest relative risk for MRI abnormalities of the acetabular labrum (p = 0.01; OR = 17) and articular cartilage (p = 0.02; OR = 5). Coxa brevis showed significant correlation and increased relative risk only for labral abnormalities (p = 0.02; OR = 6).
For hips with a normal alpha angle (Table II), coxa brevis and a higher greater trochanter showed significant association with MRI labral abnormalities only. In addition, acetabular retroversion and coxa magna also showed increased relative risk for MRI labral abnormalities, but this difference was not significant on the basis of our sample size.
The radiographic data for our ninety-nine patients (108 hips) with healed Legg-Calvé-Perthes disease showed that only nine patients (9%) had no deformity associated with labral and/or chondral abnormalities (Stulberg class I and no acetabular retroversion). Most of the patients (91%) had at least one deformity that could increase the risk for acetabular labral abnormalities. Coxa brevis was found in 74% (eighty) of all 108 hips; coxa magna, in 69% (seventy-four); an abnormal alpha angle, in 53% (fifty-seven); a higher greater trochanter, in 50% (fifty-four); and acetabular retroversion, in 50% (fifty-four).
The cases of two patients with Legg-Calvé-Perthes disease are illustrated in figures in the Appendix.
A complex variety of deformities may occur in patients with healed Legg-Calvé-Perthes disease, which might result in later hip osteoarthritis7,12. The classical mechanical concepts of coverage, containment, concentricity, and congruence have been considered the primary prognostic factors associated with hip osteoarthritis in Legg-Calvé-Perthes disease4,29,48. However, the development of preceding hip problems has recently been associated with the femoroacetabular impingement phenomenon, which is more related to repetitive motion than to axial mechanical overloading10,11. On the basis of clinical and imaging findings20, the proper diagnosis and treatment of femoroacetabular impingement may hypothetically relieve the progression of hip degeneration12,49,50.
As the hip is a ball-and-socket joint whose range of motion is limited by the osseous geometry and soft-tissue components51,52, deformities secondary to Legg-Calvé-Perthes disease may affect the joint clearance.
Femoroacetabular impingement and hinge flexion are thought to be an important source of pain in healed Legg-Calvé-Perthes disease4,12. However, patients affected by Legg-Calvé-Perthes disease can show high grades of radiographic deformity but remain almost asymptomatic with daily activities5,7,8. As the study purpose was to evaluate imaging findings, we did not take into account symptoms as a criterion of inclusion, but aimed at prospectively identifying radiographic conditions possibly related to femoroacetabular impingement, such as labral and cartilage MRI abnormalities, and abnormal head-neck morphology25. When the imaging diagnostic methods are considered, MRI is the most reasonable noninvasive method to evaluate intra-articular structures of the hip24,28 (see Appendix).
The loss of offset was the main variable related to MRI abnormalities of the acetabular labrum and cartilage. On the basis of previous studies43,44,53,54, alpha angles of <55° are considered normal for healthy individuals. Siebenrock et al.55 reported a mean alpha angle of 37° at the three o’clock position for healthy nonathlete volunteers with a mean age of seventeen years. However, a cutoff value has not been established for femoroacetabular impingement in children or for children with Legg-Calvé-Perthes disease. Despite the lack of clinical correlation, our results suggested that 55° might also be an applicable cutoff value for MRI labral abnormalities in Legg-Calvé-Perthes disease.
Nevertheless, as it was originally described, the alpha angle may not be completely feasible for all hips. For deformities like coxa plana, the center of the femoral head may be posteriorly dislocated and not coincident to the femoral neck axis in the MRI oblique axial plane or lateral radiographic view. A best-fit circle may not be perfectly achieved, and the aspherical point may be located over the articular surface and not over the cervicocapital junction. Absolute values of alpha angles with aspherical femoral heads showed low quantitative reliability, but there was qualitative agreement to assert whether the head-neck contour was normal or not.
For hips with an increased alpha angle, we found a remarkably high frequency of labral abnormalities (97%), predominantly of type III, located at the anterolateral region of the acetabular rim. However, we also found labral abnormalities on MRI scans in hips with a normal alpha angle. In these cases, our results showed that coxa brevis is the deformity most significantly associated, but an overriding greater trochanter and acetabular retroversion also trended toward an increased relative risk.
Some limitations of this study must be noted. As both observers were aware of the possibility of the diagnosis of femoroacetabular impingement, an expectative bias might have occurred. To decrease the chance of false-positive interpretations, we evaluated very carefully images suggestive of sublabral recess (see Appendix), which are considered normal variations41,42, and only abnormalities found in two or more MRI slices or planes were considered. Some images of acetabular cartilage in peripheral MRI sagittal slices of the acetabulum may actually correspond to the chondrolabral junction, and not to the articular cartilage. Therefore, a triplanar evaluation is mandatory. Nevertheless, we do not have any surgical data as a gold standard to validate our MRI findings and our comparative group is small. Although our results strongly suggest morphological and degenerative findings, several patients in our study had functional and asymptomatic hips that did not require surgical treatment. Another limitation is the lack of correlation with specific symptoms and signs of femoroacetabular impingement.
The classification system of Stulberg et al.29 could presumably identify hips prone to labral and cartilage abnormalities. However, the acetabular side has to be evaluated for the presence of acetabular retroversion. Our data showed that the Stulberg class-I group without acetabular retroversion had significant correlation with the presence of a normal labrum on MRI, whereas the class-I group with acetabular retroversion and the groups with class-II through V deformities showed no significant correlation.
In many patients with Legg-Calvé-Perthes disease, the acetabulum remodels congruently with the femoral head deformity. Lateral subluxation or extrusion of the lateral pillar due to coxa magna may also occur and, as a result, the labrum may be intermittently submitted to abnormal compression and shearing loads. Perilabral cysts may occur56. Retroversion may also occur after acetabular remodeling4,15, increasing the chance of pincer femoroacetabular impingement development. Our results did not show an increased risk for abnormalities of the Wiberg angle, but there is a fourfold risk for acetabular retroversion when the alpha angle was normal.
The high frequency of lesions located at the anterolateral portion of the acetabular rim (2h to 12h) in our study is in agreement with the data reported for patients not diagnosed with Legg-Calvé-Perthes disease35-37,41,42,57-59. A higher frequency of lesions situated at the superior portion (12h) could be explained by the enlarged cervicocapital region typically found in Legg-Calvé-Perthes disease, which is larger and wider than in patients with idiopathic femoroacetabular impingement.
In conclusion, we identified an elevated frequency of MRI abnormalities of the acetabular labrum and articular cartilage in hips with healed Legg-Calvé-Perthes disease. The loss of sphericity of the femoral head associated with a reduced offset (an abnormal alpha angle) was the most significantly associated predisposing factor for abnormalities of the acetabular labrum and cartilage, but coxa brevis with an overriding greater trochanter or coxa magna also showed an increased risk for labral abnormalities. Acetabular retroversion may predispose to labral abnormalities even when the alpha angle is normal. Hip deformities possibly related to labral and cartilage abnormalities were found in 91% of our patients.
Disclosure: None of the authors received payments or services, either directly or indirectly (i.e., via his or her institution), from a third party in support of any aspect of this work. None of the authors, or their institution(s), have had any financial relationship, in the thirty-six months prior to submission of this work, with any entity in the biomedical arena that could be perceived to influence or have the potential to influence what is written in this work. Also, no author has had any other relationships, or has engaged in any other activities, that could be perceived to influence or have the potential to influence what is written in this work. The complete Disclosures of Potential Conflicts of Interest submitted by authors are always provided with the online version of the article.