Legg-Calvé-Perthes disease is a juvenile form of osteonecrosis of the femoral head that affects 1 in 740 boys and 1 in 3,500 girls1. It remains one of the most common childhood hip disorders that can produce a permanent deformity of the femoral head and debilitating arthritis2,3. The key process involved in the pathogenesis of Legg-Calvé-Perthes disease is disruption of blood flow to the femoral head that produces osteonecrosis4. Diagnostic imaging studies demonstrate a loss of blood flow to the femoral epiphysis5-9. Histopathologic studies also demonstrate necrotic changes in the epiphyseal bone10,11. Legg-Calvé-Perthes disease is known to be a self-healing condition that advances through 4 radiographic stages: initial or osteonecrosis (Stage 1), fragmentation (Stage 2), reossification (Stage 3), and healed (Stage 4)4. One of the important elements of the healing process is the restoration of blood flow to the necrotic femoral head. Since revascularization is an essential component of necrotic bone healing that brings repair cells and soluble factors, it is important to understand the process and to determine what factors affect the rate and the quality of this process.
Gadolinium-enhanced magnetic resonance imaging (MRI) is an advanced MRI technique that depicts perfused tissues as the areas of high signal intensity. The presence of gadolinium in a perfused tissue can be enhanced by subtracting the pre-contrast images from the corresponding post-contrast images to eliminate the background signal noise9,12,13. The subtraction technique makes the areas of hypoperfusion more conspicuous due to the absence of gadolinium (seen as black areas), and it is referred to as a perfusion MRI. Studies show that a perfusion MRI is more sensitive than a noncontrast MRI in detecting and quantifying the area of hypoperfusion and bone necrosis in an early stage of Legg-Calvé-Perthes disease9,14. Preliminary studies have shown that a perfusion MRI obtained at the initial or early fragmentation stage can predict the lateral pillar involvement at the mid-fragmentation stage and the radiographic outcome at the time of a 2-year follow-up15,16. Furthermore, the amount of synovial enhancement due to hip synovitis can be quantified using a perfusion MRI17,18.
Given the utility of perfusion MRI, the current study was performed to determine the changes in the perfusion status of femoral heads affected by Legg-Calvé-Perthes disease. Our purpose was to use serial perfusion MRI to determine the pattern and rate of revascularization of the femoral heads of patients with the active stage of Legg-Calvé-Perthes disease.
Materials and Methods
Patient data were obtained from an institutional review board-approved database of prospectively enrolled patients with Legg-Calvé-Perthes disease who were assessed with a perfusion MRI between 2007 and 2014. The inclusion criterion was Legg-Calvé-Perthes disease in the initial or fragmentation stage according to the Waldenström classification19-21. Patients had anteroposterior and frog-leg lateral radiographs of the pelvis made within 6 weeks of the initial MRI to ensure correct staging. Patients were excluded if they had osteonecrosis due to trauma, corticosteroids, sickle cell disease, or other known causes. When a patient had bilateral Legg-Calvé-Perthes disease, each hip was included or excluded on the basis of its Waldenström stage. Radiographs obtained at the mid-fragmentation stage were used for the lateral pillar classification22.
Gadolinium-enhanced MRI was acquired using a GE 1.5-T scanner. Noncontrast coronal fast-spin-echo (FSE) fat-suppressed (FS) T1-weighted sequences were obtained followed by intravenous administration of 0.2 mL/kg of gadoteridol at 0.5 mmol/mL (maximum dose, 20 mL), as previously described16. Fat-suppressed coronal T1-weighted images of both hips were obtained 2 minutes after the administration of gadolinium. The subtraction technique consisted of subtracting the pre-contrast from the corresponding post-contrast coronal FSE FS T1-weighted images.
A custom-made MATLAB-based MRI analysis program called HipVasc (Texas Scottish Rite Hospital for Children, Dallas, Texas) was used to determine the percent perfusion, as previously described16. The percent perfusion of the femoral epiphysis was determined by obtaining the number of high-signal-intensity pixels (i.e., area of perfusion) on the coronal subtraction images and dividing this number by the total number of pixels in the epiphysis on the corresponding precontrast coronal images. To set the threshold for the perfused area, a black nonperfused region was selected and pixel intensities in this region were used to create an intensity threshold. Pixels with intensities above this threshold were considered perfused. Two independent observers blinded to patient information performed the measurements in order to assess the interobserver reliability of this measurement technique.
Continuous variables were first examined for normality. The Mann-Whitney U test for 2 groups and the Kruskal-Wallis test for >2 groups were used when normality was not present. A paired t test was used to compare mean perfusion percentages between the initial and follow-up MRIs. A Pearson correlation was used to examine the association of the percent perfusion with variables of interest. Interobserver reliability was assessed by calculating the intraclass correlation coefficient (ICC) with the INTRACC SAS macro (SAS). Significance was defined as a p value of <0.05.
Twenty-nine patients (30 hips) met the inclusion criteria. The average age (and standard deviation) at diagnosis was 8.4 ± 1.9 years (range, 5 to 13 years), and the average duration between the diagnosis and the first MRI was 1.1 ± 1.0 months (range, 0 to 3.6 months). All patients had at least 2 MRIs, and 21 patients (22 hips) had at least 3 (Table I). The average duration between the first and second MRIs was 5.7 ± 1.9 months. The average duration between the first and third MRIs was 10.5 ± 2.9 months. The average duration between the first and last MRIs was 13.2 ± 5.5 months. Fifteen hips were treated nonoperatively with activity modification or prolonged non-weight-bearing with the patient using crutches and/or a wheelchair. Fifteen hips were treated operatively; 10 received a femoral varus osteotomy, 4 underwent a multiple epiphyseal or femoral head-neck drilling procedure, and 1 underwent a multiple epiphyseal drilling procedure and femoral varus osteotomy.
Initial MRIs showed gadolinium enhancement in the periphery of the posterior, lateral, and medial aspects of the femoral epiphysis. Serial assessment showed a general pattern of revascularization converging toward the anterocentral region of the femoral epiphysis from the posterior, lateral, and medial aspects of the epiphysis in a horseshoe pattern. The anterocentral area of the epiphysis was the last region to show revascularization (Fig. 1). In the patients treated with the multiple epiphyseal drilling technique, gadolinium enhancement could be seen in the drill tracks in the metaphysis (Figs. 2-A and 2-B). In the series as a whole, no region of the femoral head that showed perfusion demonstrated loss of perfusion on subsequent MRIs, indicating that there was no regression of perfusion over the follow-up duration. The perfusion percentage in the necrotic epiphysis increased over the first year in all patients, but the increase was variable.
The initial MRIs showed various perfusion percentages, ranging from 5% to 70%. The average perfusion percentage (and standard deviation) on the first MRI was 35% ± 16% (Fig. 3). This increased to 77% ± 14% on the third MRI, which was acquired at an average of 10.5 ± 2.9 months later (p < 0.01). Each patient had a significantly greater gain in the perfusion percentage between the first and second MRIs (average gain, 32% ± 15%) than between the second and third MRIs (average gain, 15% ± 13%; p = 0.0003). The ICC for the percent perfusion measurements made by the 2 independent observers was 0.88 (95% confidence interval, 0.82 to 0.92).
The rate of revascularization, expressed as the change in the perfusion percentage per month, was calculated by subtracting the perfusion percentage on the first MRI from the perfusion percentage on the third MRI and dividing this value by the duration between the 2 scans. The rates were highly variable among the patients, ranging from 0.6% to 10.4% per month (Figs. 4-A through 5-B). The average rate of revascularization was 4.9% ± 2.3% per month. The areas of high signal intensity on the perfusion MRI correlated with the areas of decreased radiodensity on the corresponding radiographs (Fig. 5-A) due to resorption of the necrotic bone. The intensity of the high-signal-intensity areas decreased to the level of normal bone on follow-up MRIs as reossification occurred.
We did not find the rate of revascularization to be significantly affected by the sex of the patient (p = 0.12), age of the patient (<9 versus ≥9 years; p = 0.87), or lateral pillar involvement (A, B, or C; p = 0.42). However, the average revascularization rate was significantly higher in patients with Stage-1 disease than in those with Stage-2 (5.4% ± 2.4% versus 3.6% ± 1.1%; p = 0.032).
Currently, there is limited knowledge about the revascularization process that takes place in the necrotic femoral heads of patients with Legg-Calvé-Perthes disease. Important questions regarding this process remain unanswered and are debated; these include the source of vessel ingrowth, the pattern of revascularization, the process of repeated ischemia, and the variability in the rate of revascularization among patients. To our knowledge, this is the first study to address these questions using serial perfusion MRI. We found that revascularization was first visible at the periphery of the posterior, lateral, and medial aspects of the femoral epiphysis, with convergence toward the anterocentral region over time, which was the last area to remain avascular. Localization of gadolinium enhancement at the periphery of the necrotic epiphysis on the initial MRI suggests that the new vessels invading the necrotic epiphysis arise from the femoral neck and traverse through the margin of the articular-epiphyseal cartilage to reach the periphery. We did not observe gadolinium enhancement in the central region of the epiphysis on the initial MRIs, which argues against the possibility of metaphyseal vessels penetrating the proximal femoral physis into the epiphysis as a source of revascularization23. However, it is possible that our serial MRI follow-up duration of 13.2 ± 5.5 months was not long enough for us to observe revascularization occurring through the central region in patients with a slow revascularization process as reported by Conway7 and others24 using serial bone scintigraphy. We did observe gadolinium enhancement of the central region of the epiphysis in the patients treated with the multiple epiphyseal or femoral head-neck drilling technique, which surgically creates transphyseal channels for revascularization. It is interesting to note that the same path of new vessel ingrowth (i.e., from the femoral neck to the margin of the articular-epiphyseal cartilage to the necrotic osseous epiphysis) was observed in a piglet model of ischemic osteonecrosis25. In that animal model, expression of vascular endothelial growth factor (VEGF), a pro-angiogenic growth factor, was significantly increased in the articular-epiphyseal cartilage following ischemic osteonecrosis, stimulating revascularization and vessel invasion through the articular-epiphyseal cartilage to reach the necrotic epiphysis26.
Our study showed that the perfusion percentage increases over time but the rate varies from one patient to another. This may explain why the duration of each Waldenström stage may vary even among patients in the same age group. Each patient had a significantly greater gain in perfusion percentage between the first and second MRIs than between the second and third MRIs. The reason for the slowing down of revascularization can only be speculated at this point. It may be related to the inability of new vessels to grow past a certain distance from the periphery of the epiphysis or a loss of angiogenic stimulus in the central region of the necrotic epiphysis over time. The slowing of revascularization in the central region is consistent with the radiographic observation of this region being the last area to reossify in the reossification stage.
The slowing of revascularization of the anterocentral region of the epiphysis is consistent with previous vascular anatomy studies27,28. Chung reported that the lateral ascending cervical artery (a branch of the medial circumflex femoral artery) supplied the greatest volume of the femoral head in the pediatric population, crossing the hip capsule at the posterior trochanteric fossa27. In that study, anastomosis of the vessels at the articular cartilage-bone junction was most frequently incomplete in the anterior region, especially in males. Thus, the anterior region of the femoral head was thought to be more vulnerable to necrosis. A greater presence of blood vessels in the posterior and lateral aspects of the proximal part of the femur may explain why these regions are revascularized earlier than the anterior region, which has fewer vessels.
Revascularization of the necrotic femoral heads in patients with Legg-Calvé-Perthes disease has been studied previously with use of serial bone scintigraphy7,24. Those studies demonstrated 2 scintigraphic patterns: the A and B pathways. The A pathway began with increased radioactivity of the lateral aspect of the epiphysis followed by anterior and medial extension of the radioactivity on the scintigrams. Since bone scintigraphy does not permit accurate assessment of the medial and posterior regions of the epiphysis because of the overlying radioactivity of the acetabulum, it is unclear whether there was concomitant revascularization of the periphery of these regions in the epiphyses, as was seen in our study. In one of the previous studies, the A pathway resulted in faster revascularization, thought to be due to recanalization of existing vessels, and a good outcome24. In comparison, the B pathway began with increased radioactivity at the base of the epiphysis followed by a central extension of the radioactivity on the scintigrams. The B pathway was much slower and was associated with a worse outcome, and was thought to involve neovascularization. It is not possible to directly compare the results of our serial MRI study with the results of the scintigraphy studies as the cohorts differed in terms of patient age and disease-stage distribution, follow-up duration, and treatments. Furthermore, the imaging techniques have different capabilities, with MRI allowing higher resolution and cross-sectional imaging (i.e., multiple sections through the epiphysis in the coronal and sagittal planes). However, we do believe that the results had some similarities. In particular, the A pathway on the scintigrams resembles the revascularization process seen in our patients who had a faster rate of revascularization with early establishment of perfusion in the lateral pillar. In our previous study, patients with complete revascularization of the lateral pillar seen on MRI performed at the early stages of the disease went on to develop A or B involvement of the lateral pillar at the mid-fragmentation stage16, which predicts a good outcome. On the other hand, the B pathway appears to resemble the revascularization pattern seen in our patients who had a slow rate of revascularization. However, we did not find increased revascularization at the base of the epiphysis except in the patients treated with the multiple epiphyseal drilling technique. It is possible that our follow-up was not long enough for us to observe this process in untreated patients.
One of the theories about the pathogenesis of Legg-Calvé-Perthes disease is centered on the concept of repeated infarction of the femoral head. In contrast to this theory, we did not observe evidence of regression or fluctuation of perfusion of femoral heads on the serial MRIs. However, we cannot rule out the possibility that a repeated infarction had already taken place prior to the initial MRI or that these ischemic events occur at a later time period.
In our current clinical practice, we do not routinely obtain serial MRIs. In general, we obtain a perfusion MRI for patients >6 years old to assess the vascularity of the femoral head if they present at an early stage of the disease, when the lateral pillar classification cannot be applied. We repeat the MRI for patients when we need an accurate assessment of femoral head revascularization and healing to determine if it is safe for them to start full weight-bearing. It is important to note that generally our recommendation of full weight-bearing is not based on the presence of revascularization alone (i.e., high signal intensity) on perfusion MRI but is based on normalization of the signal intensity of the revascularized area on perfusion MRI, especially the lateral pillar region, due to reossification, as shown in Figure 5-A. When a patient has been treated with the multiple epiphyseal drilling technique, we repeat the MRI to assess the healing response to the treatment and to determine if there is a persistent avascular region that needs to be treated with repeat drilling.
It should be noted that serial administration of gadolinium contrast agents has been shown to be associated with gadolinium deposition in the brain29,30. While it is not yet known if this has clinical implications, the need for serial perfusion MRI should be carefully considered case by case. Of the 2 types of gadolinium contrast agents available commercially (linear and macrocyclic), linear agents are associated with brain deposition as they are more unstable29. Thus, it is preferable to utilize macrocyclic agents, such as the gadoteridol used in this study, for perfusion MRI.
This study has some limitations. The sample size was small because assessment of femoral heads affected by Legg-Calvé-Perthes disease represents a relatively new application of perfusion MRI and because we limited the study to patients with the early stages of the disease. Another limitation is that the study included both patients treated nonoperatively and those treated with operative means and did not include a control group that truly represented the natural history of the disease. Lastly, the duration between the MRIs and the number of MRIs obtained per patient were not rigidly controlled because of patient availability, physician preference, and MRI scheduling. Twenty-one of the 29 patients, however, had ≥3 perfusion MRIs, and most scans were performed about 5 months apart. The strengths of the study include prospective enrollment of the patients, assessment of the MRIs by 2 independent observers, and all patients having the active stage of Legg-Calvé-Perthes disease.
In summary, revascularization of the necrotic femoral head in the active stage of Legg-Calvé-Perthes disease increased over time, but the rate of revascularization was highly variable among patients. A horseshoe pattern of revascularization was seen in the periphery of the posterior, lateral, and medial aspects of the femoral epiphysis, which converged toward the anterocentral region.
Investigation performed at the Center of Excellence in Hip Disorders, Texas Scottish Rite Hospital for Children, Dallas, Texas
A commentary by Pablo Castañeda, MD, is linked to the online version of this article at jbjs.org.
Disclosure: There was no external funding source. The Disclosure of Potential Conflicts of Interest forms are provided with the online version of the article.
1. Molloy MK, MacMahon B. Incidence of Legg-Perthes disease (osteochondritis deformans). N Engl J Med. 1966 ;275(18):988–90.
2. McAndrew MP, Weinstein SL. A long-term follow-up of Legg-Calvé-Perthes disease. J Bone Joint Surg Am. 1984 ;66(6):860–9.
3. Larson AN, Sucato DJ, Herring JA, Adolfsen SE, Kelly DM, Martus JE, Lovejoy JF, Browne R, Delarocha A. A prospective multicenter study of Legg-Calvé-Perthes disease: functional and radiographic outcomes of nonoperative treatment at a mean follow-up of twenty years. J Bone Joint Surg Am. 2012 ;94(7):584–92.
4. Kim HK, Herring JA. Pathophysiology, classifications, and natural history of Perthes disease. Orthop Clin North Am. 2011 ;42(3):285–95, v.
5. Kim HK. Legg-Calve-Perthes disease: etiology, pathogenesis, and biology. J Pediatr Orthop. 2011 ;31(2)(Suppl):S141–6.
6. Atsumi T, Yamano K, Muraki M, Yoshihara S, Kajihara T. The blood supply of the lateral epiphyseal arteries in Perthes’ disease. J Bone Joint Surg Br. 2000 ;82(3):392–8.
7. Conway JJ. A scintigraphic classification of Legg-Calvé-Perthes disease. Semin Nucl Med. 1993 ;23(4):274–95.
8. de Camargo FP, de Godoy RM Jr, Tovo R. Angiography in Perthes’ disease. Clin Orthop Relat Res. 1984 ;191:216–20.
9. Lamer S, Dorgeret S, Khairouni A, Mazda K, Brillet PY, Bacheville E, Bloch J, Penneçot GF, Hassan M, Sebag GH. Femoral head vascularisation in Legg-Calvé-Perthes disease: comparison of dynamic gadolinium-enhanced subtraction MRI with bone scintigraphy. Pediatr Radiol. 2002 ;32(8):580–5. Epub 2002 Jun 14.
10. Catterall A, Pringle J, Byers PD, Fulford GE, Kemp HB, Dolman CL, Bell HM, McKibbin B, Rális Z, Jensen OM, Lauritzen J, Ponseti IV, Ogden J. A review of the morphology of Perthes’ disease. J Bone Joint Surg Br. 1982;64(3):269–75.
11. Jonsater S. Coxa plana; a histo-pathologic and arthrografic study. Acta Orthop Scand Suppl. 1953;12:5–98.
12. Mahnken AH, Staatz G, Ihme N, Günther RW. MR signal intensity characteristics in Legg-Calvé-Perthes disease. Value of fat-suppressed (STIR) images and contrast-enhanced T1-weighted images. Acta Radiol. 2002 ;43(3):329–35.
13. Sebag G, Ducou Le Pointe H, Klein I, Maiza D, Mazda K, Bensahel H, Hassan M. Dynamic gadolinium-enhanced subtraction MR imaging—a simple technique for the early diagnosis of Legg-Calvé-Perthes disease: preliminary results. Pediatr Radiol. 1997 ;27(3):216–20.
14. Kim HK, Kaste S, Dempsey M, Wilkes D. A comparison of non-contrast and contrast-enhanced MRI in the initial stage of Legg-Calvé-Perthes disease. Pediatr Radiol. 2013 ;43(9):1166–73. Epub 2013 Mar 12.
15. Du J, Lu A, Dempsey M, Herring JA, Kim HK. MR perfusion index as a quantitative method of evaluating epiphyseal perfusion in Legg-Calve-Perthes disease and correlation with short-term radiographic outcome: a preliminary study. J Pediatr Orthop. 2013 ;33(7):707–13.
16. Kim HK, Wiesman KD, Kulkarni V, Burgess J, Chen E, Brabham C, Ikram H, Du J, Lu A, Kulkarni AV, Dempsey M, Herring JA. Perfusion MRI in early stage of Legg-Calvé-Perthes disease to predict lateral pillar involvement: a preliminary study. J Bone Joint Surg Am. 2014 ;96(14):1152–60. Epub 2014 Jul 16.
17. Kamiya N, Yamaguchi R, Adapala NS, Chen E, Neal D, Jack O, Thoveson A, Gudmundsson P, Brabham C, Aruwajoye O, Drissi H, Kim HK. Legg-Calvé-Perthes disease produces chronic hip synovitis and elevation of interleukin-6 in the synovial fluid. J Bone Miner Res. 2015 ;30(6):1009–13.
18. Neal DC, O’Brien JC, Burgess J, Jo C, Kim HK; International Perthes Study Group. Quantitative assessment of synovitis in Legg-Calvé-Perthes disease using gadolinium-enhanced MRI. J Pediatr Orthop B. 2015 ;24(2):89–94.
19. Joseph B, Varghese G, Mulpuri K, Narasimha Rao K, Nair NS. Natural evolution of Perthes disease: a study of 610 children under 12 years of age at disease onset. J Pediatr Orthop. 2003 ;23(5):590–600.
20. Waldenström H. The definitive form of the coxa plana. Acta Radiol. 1922 ;1:384–94.
21. Hyman JE, Trupia EP, Wright ML, Matsumoto H, Jo CH, Mulpuri K, Joseph B, Kim HK; International Perthes Study Group Members. Interobserver and intraobserver reliability of the modified Waldenström classification system for staging of Legg-Calvé-Perthes disease. J Bone Joint Surg Am. 2015 ;97(8):643–50.
22. Herring JA, Neustadt JB, Williams JJ, Early JS, Browne RH. The lateral pillar classification of Legg-Calvé-Perthes disease. J Pediatr Orthop. 1992 ;12(2):143–50.
23. Bohr HH. On the development and course of Legg-Calvé-Perthes disease (LCPD). Clin Orthop Relat Res. 1980 ;150:30–5.
24. Tsao AK, Dias LS, Conway JJ, Straka P. The prognostic value and significance of serial bone scintigraphy in Legg-Calvé-Perthes disease. J Pediatr Orthop. 1997 ;17(2):230–9.
25. Kim HK, Bian H, Aya-ay J, Garces A, Morgan EF, Gilbert SR. Hypoxia and HIF-1alpha expression in the epiphyseal cartilage following ischemic injury to the immature femoral head. Bone. 2009 ;45(2):280–8. Epub 2009 Apr 5.
26. Kim HK, Bian H, Randall T, Garces A, Gerstenfeld LC, Einhorn TA. Increased VEGF expression in the epiphyseal cartilage after ischemic necrosis of the capital femoral epiphysis. J Bone Miner Res. 2004 ;19(12):2041–8. Epub 2004 Sep 20.
27. Chung SM. The arterial supply of the developing proximal end of the human femur. J Bone Joint Surg Am. 1976 ;58(7):961–70.
28. Trueta J. The normal vascular anatomy of the human femoral head during growth. J Bone Joint Surg Br. 1957 ;39-B(2):358–94.
29. Malayeri AABK, Brooks KM, Bryant LH, Evers R, Kumar P, Reich DS, Bluemke DA. National Institutes of Health perspective on reports of gadolinium deposition in the brain. J Am Coll Radiol. 2016 ;13(3):237–41. Epub 2016 Jan 23.
30. Ramalho J, Semelka RC, Ramalho M, Nunes RH, AlObaidy M, Castillo M. Gadolinium-based contrast agent accumulation and toxicity: an update. AJNR Am J Neuroradiol. 2015 . Epub 2015 Dec 10.