The cerebral palsy hip classification is reliable AN INTER- AND INTRA-OBSERVER RELIABILITY STUDY
M. L. Murnaghan, P. Simpson, J. G. Robin, B. J. Shore, P. Selber, H. K. Graham From the Royal Children’s Hospital, Parkville, Australia
We have tested the reliability of a recently reported classification system of hip morphology in adolescents with cerebral palsy in whom the triradiate cartilage was closed. The classification is a six-grade ordinal scale, based on the measurement of the migration percentage and an assessment of Shenton’s arch, deformity of the femoral head, acetabular deformity and pelvic obliquity. Four paediatric orthopaedic surgeons and four physiotherapists received training in the use of the classification which they applied to the assessment of 42 hip radiographs, read on two separate occasions. The inter- and intra-observer reliability was assessed using the intraclass correlation coefficient and found to be excellent, with it ranging from 0.88 to 0.94. The classification in our study was shown to be valid (based on migration percentage), and reliable. As a result we believe that it can now be used in studies describing the natural history of hip displacement in cerebral palsy, in outcome studies and in communication between clinicians.
©2010 British Editorial Society of Bone and Joint Surgery doi:10.1302/0301-620X.92B3. 23105 $2.00
Population-based studies have shown that displacement of the hip affects approximately onethird of children with cerebral palsy1-3 and has been directly related to gross motor function as determined by the Gross Motor Function Classification System (GMFCS).4 A wide variety of treatments are used in the management of this type of hip displacement ranging from bracing and botulinium injection to surgical reconstruction and salvage surgery. Understanding the natural history and clinical outcome is hampered by the lack of a valid and reliable classification system. In children with developmental dysplasia of the hip (DDH), the Severin5 classification is widely used to describe the outcome of treatment. However, the inter- and intraobserver reliability of this classification system has been questioned.6 In a recent study, a new system describing hip morphology in adolescents with cerebral palsy was proposed.7 It uses a six-grade ordinal scale based on the measurement of hip migration percentage8 and assessment of the integrity of Shenton’s arch, deformity of the femoral head, acetabular deformity and pelvic obliquity. Our aim in this study was to test the inter- and intraobserver reliability of this system.
J Bone Joint Surg [Br] 2010;92-B:436-41. Received 17 July 2009; Accepted after revision 6 November 2009
Patients and Methods Four orthopaedic surgeons (JGR, BJS, PS, AK) and four physiotherapists (AF, TH-I, JR, PT)
M. L. Murnaghan, MD, MEd, FRCS (C), Orthopaedic Fellow B. J. Shore, MD, FRCS (C), Orthopaedic Fellow P. Selber, MD, FRACS, Orthopaedic Surgeon H. K. Graham, MD, FRCS(Ed), FRACS, Professor of Orthopaedic Surgery Hugh Williamson Gait Laboratory, Royal Children’s Hospital, Flemington Road, Parkville, Victoria 3052, Australia. J. G. Robin, MB BS, Orthopaedic Registrar Murdoch Childrens Research Institute, Flemington Road, Parkville, Victoria 3052, Australia. P. Simpson, BSc (Hons), Statistician Monash University, Commercial Road, Melbourne, Victoria 3004, Australia. Correspondence should be sent to Professor H. Kerr Graham; email:
[email protected]
436
were instructed in the grading of hip radiographs using the classification system.7 Two of the surgeons were fellowship trained in paediatric orthopaedic surgery and had been in practice for a mean of six years (4 to 10). The other two surgeons had completed their orthopaedic residencies and were in the final months of their paediatric orthopaedic fellowship. The four physiotherapists had a mean clinical experience of 20 years (16 to 26) and were actively involved in research and clinical work with patients with cerebral palsy. All assessors were introduced to the classification system by an oral presentation and given with written instructions and guidelines. They were given printed radiographs as well as a pen or pencil, ruler and calculator. The hip classification. The following six grades as originally described by Robin et al7 are described below and illustrated in Figure 1. Grade I: normal hip . A morphologically normal hip at skeletal maturity has a migration percentage < 10%. Shenton’s arch is intact and the femoral head is round and is covered by a welldeveloped acetabulum. This includes an everted, well-defined lateral acetabular margin, and a normal teardrop and sourcil. Pelvic obliquity is < 10°. Grade II: nearly normal hip. A nearly normal hip has a migration percentage ≥ 10% and ≤ 15%. Shenton’s arch is intact and the femoral head is round or almost round. The THE JOURNAL OF BONE AND JOINT SURGERY
THE CEREBRAL PALSY HIP CLASSIFICATION IS RELIABLE
437
Fig. 1 Radiographs and diagrams showing the hip classification system in cerebral palsy. Reproduced with permission from Robin et al.7
acetabulum may be normal or only slightly deformed with the lateral acetabular margin slightly blunted. The teardrop is slightly widened, but the sourcil should be well formed. Pelvic obliquity is < 10°. Grade III: dysplastic hip. A dysplastic hip has a migration percentage > 15% and ≤ 30%. Shenton’s arch is intact or broken by ≤ 5 mm. The femoral head may be round or VOL. 92-B, No. 3, MARCH 2010
slightly flattened. The acetabulum is normal or mildly dysplastic with blunt lateral acetabular margin or a gothic arch. There is a widened teardrop and a poorly developed sourcil. Pelvic obliquity is < 10°. Grade IV: subluxed hip. A subluxed hip has a migration percentage > 30% and < 100%. There is always some contact of the femoral head with the true acetabulum. Shenton’s arch is
438
M. L. MURNAGHAN, P. SIMPSON, J. G. ROBIN, B. J. SHORE, P. SELBER, H. K. GRAHAM
Table I. Details of intra-observer reliability Rater
ICC
*
1 0.92 2 0.88 3 0.94 4 0.91 5 0.94 6 0.90 7 0.93 8 0.88 * ICC, intraclass correlation coefficient
Table II. Details of inter-observer reliability
95% confidence interval (CI)
Rater
0.87 to 0.97 0.82 to 0.95 0.90 to 0.97 0.86 to 0.96 0.90 to 0.98 0.85 to 0.96 0.87 to 0.97 0.87 to 0.95
All raters Reading A Reading B Physiotherapists Reading A Reading B Orthopaedic surgeons Reading A Reading B
ICC*
95% CI†
0.85 0.91
0.79 to 0.91 0.87 to 0.95
0.84 0.92
0.77 to 0.91 0.88 to 0.96
0.88 0.89
0.83 to 0.94 0.84 to 0.94
* ICC, intraclass correlation coefficient † CI, confidence interval
broken by more than 5 mm. The femoral head has variable deformity from none to severe with more than half being affected. The acetabulum likewise has variable deformity from being normally shaped to having a large gothic arch. Grade V: dislocated hip. A dislocated hip has a migration percentage ≥ 100%. There is no remaining contact between the femoral head and the true acetabulum. Shenton’s arch is completely disrupted. As in grade IV, there is variable deformity of the femoral head and acetabulum (relating to the timing and rate of dislocation) with variable degrees of pelvic obliquity. Grade VI: salvage surgery - loss of hip. These hips have undergone salvage surgery because of painful dislocation or subluxation. Salvage surgery may involve valgus osteotomy alone, excision arthroplasty plus or minus valgus osteotomy, arthrodesis or replacement arthroplasty. Radiographs of 42 hips of adolescents with cerebral palsy aged between 14 and 19 years and in whom the triradiate cartilage was closed were drawn from a large, population-based series of children with cerebral palsy born between January 1990 and December 1992 from the cerebral palsy register of the State of Victoria, Australia.9 The radiographs were selected by the first author (MLM). He had previously classified the radiographs based on the published classification system.7 Drawing from approximately 80 possible radiographs, 42 were selected in a nonrandomised way to provide a balanced representation of all six grades (grade I, 3 hips; grade II, 10 hips; grade III, 6 hips; grade IV, 8 hips; grade V, 9 hips; grade VI, 6 hips). The radiographs were taken using a standardised positioning protocol employed in the Radiology Department, which has been shown to improve reliability.10 They were read on two separate occasions by all eight readers. Each reader was asked to classify each of the hips into one of six grades as defined by the classification system. All identifying information was removed from the radiographs which were presented in two sessions in a different order and separated by four weeks. Statistical analysis. This was undertaken using the intraclass correlation coefficient. This is equivalent to the weighted kappa value in which less weight is assigned to agreement since categories are further apart.11 The intraclass correlation coefficient was interpreted using established conven-
tions for kappa in which < 0 is poor agreement, 0 to 0.2 slight, 0.2 to 0.4 fair, 0.4 to 0.6 moderate, 0.6 to 0.8 substantial, and > 0.8 almost perfect agreement.12 In addition to inter- and intra-observer reliability, comparison was also made with an established standard of consensus grade agreed upon by the first and senior authors (MLM, HKG) who were blinded to the results of the eight readers. An asymptotic symmetry test was used to test for bias, in order to determine if one rating was consistently higher or lower than the other. All statistical analysis was performed using Stata 10.0 (StataCorp, College Station, Texas).13
Results Intra-observer reliability. The results are summarised in
Table I. The intraclass correlation coefficients ranged from 0.88 to 0.94 demonstrating excellent agreement. A symmetry test indicated that there was no bias when comparing the first and second readings of any of the raters. Inter-observer reliability. The results are summarised in Table II. Having established the intra-observer reliability on their first and second reading, the data were analysed for interobserver reliability. The reliability was excellent for both readings, with the mean intraclass correlation coefficient increasing slightly between the two readings (0.84 to 0.88 and 0.89 to 0.92, respectively). When taken individually, the physiotherapists showed excellent reliability, with noted improvement (although this did not reach statistical significance in the intraclass correlation coefficent between the two readings (0.84 and 0.92)). The orthopaedic surgeons were similarly excellent in terms of reliability (0.88 and 0.89). Agreement with the established standards. The data were also analysed for agreement with the established standard (Table III). Agreement with the established standard was excellent with a mean intraclass correlation coefficient of 0.91 (0.74 to 0.96). Of 16 readings, 15 scored an intraclass correlation coefficient greater than 0.85, with one reader recording an intraclass correlation coefficient of 0.74 which is the only reading not in the ‘almost perfect’ category when compared with the established standard. Patterns of disagreement. The data were further analysed to determine if there were individual subjects (radiographs) THE JOURNAL OF BONE AND JOINT SURGERY
THE CEREBRAL PALSY HIP CLASSIFICATION IS RELIABLE
Table III. Details of reader agreement with the established standards ICC (95% CI)*
ICC (95% CI)
Rater
Reading A
Reading B
1 2 3 4 5 6 7 8
0.94 0.92 0.94 0.86 0.96 0.91 0.90 0.74
0.92 0.93 0.93 0.87 0.95 0.96 0.93 0.89
(0.90 to 0.97) (0.87 to 0.97) (0.90 to 0.98) (0.78 to 0.94) (0.94 to 0.98) (0.86 to 0.96) (0.85 to 0.96) (0.60 to 0.88)
(0.87 to 0.97) (0.88 to 0.97) (0.89 to 0.97) (0.82 to 0.95) (0.92 to 0.98) (0.93 to 0.98) (0.89 to 0.97) (0.83 to 0.96)
* ICC, intraclass correlation coefficient; CI confidence interval)
Table IV. Details of patterns of disagreement (individual subjects (radiographs) which had a difference in grade greater than 2 between their ratings). Gradings according to the classification of Robin et al7 Subject
Gradings
Number of readers
12 14 16 20 25 38
II, V II, IV IV, VI II, IV IV, VI II, IV
1 3 4 1 1 1
which consistently demonstrated poorer agreement. Table IV highlights those which led to a difference of more than two points on the hip scale which was assigned. Radiographs 14 and 16 showed repeated disagreement between readers, three and four times, respectively. A similar analysis was carried out on subjects which had been consistently rated differently by readers and the established standard. Table V summarises these results. In all but one case, the assigned grade by the readers was higher than the established standard.
Discussion The migration percentage as described by Reimers8 is the most widely used measurement of hip displacement in children with cerebral palsy. In his original report he suggested that measurement of the migration percentage was more reliable than the use of the centre-edge angle in spastic subluxation of the hip. The centre-edge angle was found to be more variable depending on the position of the child when the radiograph was taken and to be non-linear. He estimated an SEM of 10% for the migration percentage based on an estimate of the error involved in measuring a line to the nearest millimetre. In addition, two groups have studied the reliability of the measurement of the migration percentage. Using carefully standardised methodology, Parrott et al14 found that an experienced reader would be expected to measure the migration percentage on a single radiograph to within 5.8% of the true value and hence a change in migraVOL. 92-B, No. 3, MARCH 2010
439
Table V. Details of patterns of disagreement (individual subjects (radiographs) which had a difference in grade > 2 between their rating (either A or B) and the established standard. Gradings according to the classification of Robin et al7 Subject
Grade
Established standard
Number of times
1 12 14 16 20 25 38
III V IV or VI VI IV or V IV or VI I
I II II IV II II III
3 1 4 4 3 3 1
tion percentage between two radiographs taken at different times to within 8.3% of the true value. More recently, Faraj, Atherton and Stott15 found that a change of more than 13% in the migration percentage, in repeated measurements by one assessor, was required in order to be 95% confident of a ‘true’ change. However, both of these studies examined only a single radiograph. In a clinical setting, most management decisions are based on the examination of a series of radiographs taken at different times. Advantages of the use of the migration percentage are that it is easily understood, has good face validity, acceptable reliability and there is an extensive body of literature which has related specific values of this important clinical outcomes.1-3,8,10,14-17 For example, a threshold value of 40% has been shown to be associated with instability and a rapid increase in the migration percentage in most hips studied.16 Most definitions of subluxation and dislocation as well as thresholds for intervention and management decisions, are commonly based on the migration percentage. A further advantage is that it can be measured from a good quality anteroposterior radiograph of the hip once the ossific nucleus is present. Measurement of the migration percentage from serial hip radiographs has also been proposed as the most practical method of carrying out hip surveillance programmes in children with cerebral palsy.17 Given the frequency of silent hip displacement in children with cerebral palsy, a number of authors have proposed systematic hip surveillance for the early detection of hip displacement.10,17 It is therefore widely used for the detection of early hip displacement, the longitudinal follow-up of children who have been identified with hip displacement and for monitoring the outcome of various interventions.18-20 There are inherent limitations to the measurement of the migration percentage from a twodimensional anteroposterior radiograph of the hip. Dislocation of the hip is usually posterolateral in children with cerebral palsy, but is occasionally anterior.21 Both anterior and posterior dislocations do not always show a marked increase in the migration percentage although there is almost always some degree of lateralisation before the femoral head escapes from the acetabulum. Axial imaging using CT com-
440
M. L. MURNAGHAN, P. SIMPSON, J. G. ROBIN, B. J. SHORE, P. SELBER, H. K. GRAHAM
Fig. 2
Fig. 3
Anteroposterior radiograph of the hips of a 15-year-old boy with left hemiplegia, gross motor function classification system level II and painful, neglected left hip subluxation. The right hip is grade I and the left hip is grade IV. The deformity of the femoral head (left hip) is moderate with lateral depression, the acetabulum is dysplastic with a small gothic arch and pelvic obliquity is mild, measuring 15°.
Anteroposterior radiograph of the hips of a 17-year-old boy with spastic quadriplegia, gross motor function classification system (GMFCS) level IV. The hips were managed by adductor releases, bilateral femoral osteotomies and removal of the implants. At skeletal maturity he has no pain and comfortable sitting. The right hip is grade III (migration percentage (MP) 28%) and the left also grade III (MP 20%). Despite reconstructive surgery, there are few normal hips at skeletal maturity in GMFCS levels IV and V.
plemented by three-dimensional reconstruction, and MRI offer more detailed information.22,23 However, although CT may be useful when planning reconstruction of the hip in an individual patient, neither CT nor MRI is appropriate for the classification of hips in population-based studies or in repeated longitudinal examinations in a single subject.24 CT examination of the hips exposes children to a significant dose of ionising radiation and MRI usually requires general anaesthesia.24 Given the widespread use of the migration percentage in studies of hip displacement in children with cerebral palsy, we considered that this was the key continuous variable to be used in the classification of the hip in such patients. However, with the development of severe displacement, various combinations of deformity of the femoral head, acetabular deformity (gothic arch) and pelvic obliquity make the measurement of the migration percentage increasingly difficult. The hip classification system was therefore designed to describe the morphology of hips after skeletal maturity in cerebral palsy using a combination of both quantitative (migration percentage) and qualitative features. The classification, based on the use of a continuous variable migration percentage was consciously based on the GMFCS4 for children with cerebral palsy and its relationship to the gross motor function measure.25 The study by Ward et al6 of the Severin system for DDH found unacceptably low intra- and inter-observer reliability. This re-inforces the importance of establishing the validity and reliability of a new classification system before its widespread acceptance and application. Previous studies
have discussed the influence of the expertise of the readers on interobserver agreement.26,27 In our study, both physiotherapists and orthopaedic surgeons served as readers. There was a spectrum of clinical experience in both groups. There was a subtle increase in the interobserver reliability between the first (A) and second (B) reading which could be attributed to improved application of the system secondary to learning from the experience of the first reading. In order to minimise the influence of recall bias, the radiographs were presented four weeks apart in addition to being blinded and presented in a different order. In our study the classification has been shown to be valid and reliable when used by experienced physiotherapists and paediatric orthopaedic surgeons. As a result it can now be used to describe the natural history and outcome of hip displacement in cerebral palsy. Basing the classification on the migration percentage allows comparison with studies reporting a change in the migration percentage in younger children.19,20 However, these rarely show the morphological features which we found to be part of hip disease in adolescents with cerebral palsy. We believe that the changes in hip morphology accelerate during the adolescent growth spurt. Although the rate of change may slow with the onset of skeletal maturity, it is known that scoliosis, pelvic obliquity and changes in hip status may occur in the third and fourth decades.28 The classification may be useful as a means of communication regarding the clinical status of hip morphology in adolescents with cerebral palsy as well as a tool for evaluating the THE JOURNAL OF BONE AND JOINT SURGERY
THE CEREBRAL PALSY HIP CLASSIFICATION IS RELIABLE
441
References 1. Soo B, Howard JJ, Boyd RN, et al. Hip displacement in cerebral palsy. J Bone Joint Surg [Am] 2006;88-A:121-9. 2. Hagglund G, Lauge-Pederson H, Wagner P. Characteristics of children with hip displacement in cerebral palsy. BMC Musculoskeletal Disord 2007;8:101. 3. Connelly A, Flett P, Graham HK, Oates J. Hip surveillance in Tasmanian children with cerebral palsy. J Paediatr Child Health 2009;45:437-43. 4. Palisano R, Rosenbaum P, Walter S, et al. Development and reliability of a system to classify gross motor function in children with cerebral palsy. Dev Med Child Neurol 1997;39:214-23. 5. Severin E. Contribution to the knowledge of congenital dislocation of the hip joint. Acta Chirurgica Scandinavica 1941;84(Suppl 63):37-54. 6. Ward W, Vogt M, Crucziak JS, et al. Severin classification system for evaluation of the results of operative treatment of congenital dislocation of the hip: a study of intraobserver and interobserver reliability. J Bone Joint Surg [Am] 1997;79-A:656-63. 7. Robin J, Graham HK, Baker R, et al. A classification system for hip disease in cerebral palsy. Dev Med Child Neurol 2009;51:183-92. 8. Reimers J. The stability of the hip in children: a radiological study of results of muscle surgery in cerebral palsy. Acta Orthop Scand 1980;184:1-100. Fig. 4 Anteroposterior radiograph of the hips of a 16-year-old girl with spastic quadriplegia, gross motor function classification system level V. The right hip (grade VI) has been excised for intractable pain. However, there has been proximal migration with contact between the femur and pelvis causing pain. The left hip is grade II with a mildly flat femoral head which is completely covered (migration percentage 0%) and there is moderate pelvic obliquity of 30°.
9. Howard J, Soo B, Graham HK, et al. Cerebral palsy in Victoria: motor types, topography and gross motor function. J Pediatr Child Health 2005;41:479-83. 10. Dobson F, Boyd RN, Parrott J, Nattrass GR, Graham HK. Hip surveillance in children with cerebral palsy: impact on the surgical management of spastic hip disease. J Bone Joint Surg [Br] 2002;84-B:720-6. 11. Fleiss JL, Cohen J. The equivalence of weighted kappa and the intraclass correlation coefficient as measures of reliability. Educational and Psychological Measurement 1973;33:613-19. 12. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33:159-74. 13. StataCorp. 2007. Stata Statistical Software: Release 10. College Stations. TX: StataCorp LP.
outcome of treatment and natural history studies. The combined use of the GMFCS and the classification may also be useful. For example, adolescents at GMFCS level I and II, who are long-term independent community walkers may require a normal or near morphologically normal hip (Fig. 2). Those with lower levels of mobility or who are non-walkers may be able to function with less normal hip development (Fig. 3). Grade-VI hips have been included in the classification to allow the documentation of hips which have been painful and required excision as opposed to those which are dislocated (grade V), but have not yet required salvage (Fig. 4). The relationship between hip morphology, pain and sitting and standing function deserves further exploration. Our results show that the hip scale in cerebral palsy is a reliable tool for the evaluation of radiological findings of hip morphology in adolescents with cerebral palsy in whom the triradiate cartilage is closed. The interobserver reliability was excellent, showing consistent agreement between readings for individual raters.
Supplementary material Appendices relating to the classification of deformity of the femoral head, acetabular deformity and pelvic obliquity are available with the electronic version of this article on our website at www.jbjs.org.uk We wish to thank A. Fosang, T. Hastings-Ison, J. Rodda and P. Thomason for their help in grading the radiographs. We also acknowledge the help of M. Sheedy in the preparation of this manuscript. No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article. VOL. 92-B, No. 3, MARCH 2010
14. Parrott J, Boyd RN, Dobson F, et al. Hip displacement in spastic cerebral palsy: repeatability of radiological measurement. J Pediatr Orthop 2002;22:660-7. 15. Faraj S, Atherton WG, Stott NS. Inter- and intra-measurer error in the measurement of Reimers’ hip migration percentage. J Bone Joint Surg [Br] 2004;86-B:4347. 16. Hagglund G, Lauge-Pederson H, Persson M. Radiographic threshold values for hip screening in cerebral palsy. J Child Orthop 2007;1:43-7. 17. Scrutton D, Baird G. Surveillance measures of the hip of children with bilateral cerebral palsy. Arch Dis Childhood 1997;76:381-41. 18. Graham HK, Boyd R, Carlin JB, et al. Does Botulinum toxin a combined with hip bracing prevent hip displacement in childen with cerebral palsy and “hips-atrisk”?: a randomized controlled trial. J Bone Joint Surg [Am] 2008;90-A:23-33. 19. Khot A, Sloan S, Desai S, et al. Adductor release and chemodenervation in children with cerebral palsy: a pilot study in 16 children. J Child Orthop 2008;2:293-9. 20. Miller F, Cardoso Dias R, Dabny KW, Lipton GE, Triana M. Soft-tissue release for spastic hip subluxation in cerebral palsy. J Pediatr Orthop 1997;17:571-84. 21. Selva G, Miller F, Dabney KW. Anterior hip dislocation in children with cerebral palsy. J Pediatr Orthop 1998;18:54-61. 22. Kim HT, Wenger DR. Location of acetabular deficiency and associated hip dislocation in neuromuscular hip dysplasia: three-dimensional computed tomographic analysis. J Pediatr Orthop 1994;17:143-51. 23. Chambers HG. A classification for hip disease in cerebral palsy. Dev Med Child Neurol 2009;51:168-9. 24. Robin J, Graham HK, Selber P, et al. Proximal femoral geometry in cerberal palsy: a population-based cross-sectional study. J Bone Joint Surg [Br] 2008;90B:1372-9. 25. Hanna SE, Rosenbaum PL, Bartlett DJ, et al. Stability and decline in gross motor function among children and youth with cerebral palsy aged 2 to 21 years. Dev Med Child Neurol 2009;51:295-302. 26. Kristiansen B, Andersen UL, Olsen CA, Vanmarken JE. The Neer classification of fractures of the proximal humerus: an assessment of interobserver variation. Skeletal Radiol 1988;17:420-2. 27. Rasmussen S, Madsen PV, Bennicke K. Observer variation in the LaugeHansen classification of ankle fractures: precision improved by instruction. Acta Orthop Scand 1993;64:693-4. 28. Miller F. Cerebral palsy. Section 1. Chapter 10. New York: Springer, 2005:523666.