THE EFFECT OF A TECHNICAL QUALITY ASSESSMENT OF HIP-EXTENDED RADIOGRAPHS ON INTEROBSERVER AGREEMENT IN THE DIAGNOSIS OF CANINE HIP DYSPLASIA

THE EFFECT OF A TECHNICAL QUALITY ASSESSMENT OF HIP-EXTENDED RADIOGRAPHS ON INTEROBSERVER AGREEMENT IN THE DIAGNOSIS OF CANINE HIP DYSPLASIA GEERT E.C. VERHOEVEN, RUTH R. FORTRIE, LUC DUCHATEAU, JIMMY H. SAUNDERS, BERNADETTE VAN RYSSEN, HENRI VAN BREE, FRANK COOPMAN Experienced and inexperienced observers evaluated the assessability of 50 radiographs (25 dogs) and determined the hip status (dysplasia/nondysplasia and final scoring according Fe dération Cynologique Internationale [FCI]- criteria) individually. A radiographic technical quality assessment was performed in a separate reading session. Interobserver agreement in determining dysplasia/nondysplasia and FCI-scoring did not significantly increase with the increasing quality of a radiograph, irrespective whether these observers are experienced or not. There was a significant agreement between the technical quality assessment and assessability (Po0.0005). Despite the effort to objectify radiographic quality and to present high-quality radiographs to observers, interobserver agreement on dysplasia/nondysplasia and final scoring, remains low, even in the experienced group. Although increased radiographic quality narrows the range of scoring, the range remains unacceptably high. r 2010 Veterinary Radiology & Ultrasound, Vol. 51, No. 5, 2010, pp 498 503. Key words: agreement, dog, hip dysplasia. Introduction ROGRESS IN DECREASING the incidence of canine hip Pdysplasia (CHD) remains low 1,2 despite the extensive use of various protocols using the subjective hip-extended method to screen breeding dogs. One of the reasons for this slow progress may be the inability to define a dysplastic dog or to define a dog s classification as A, B, C, D, or E according to the FCI-scoring system in an objective, repeatable, and correct way. The correct CHD diagnosis is pivotal to making sound recommendations regarding breeding programs to reduce the frequency of CHD and may prove useful in the clinical management and treatment of CHD. 3 9,2,10 14 The CHD diagnosis is characterized by low interobserver agreement for final scoring, which could be due to improper positioning of the pelvis and/or femurs, or to insufficient radiographic contrast and/or detail. The importance of radiographic quality in the diagnosis of CHD has been described. 15 20 From the Department of Medical Imaging (Verhoeven, Saunders, Van Ryssen, Van Bree, Coopman), and the Department of Physiology and Biometrics (Duchateau), Faculty of Veterinary Medicine, Ghent University, Salisburylaan 133, 9820 Merelbeke, Belgium, the Algemene Dierenkliniek Randstad, Frans Beirenslaan 155, 2150 Borsbeek, Belgium (Verhoeven, Fortrie), and the Department of Biosciences and Landscape Architecture, University College Ghent, Voskenslaan 290, 9000 Gent, Belgium (Coopman). Address correspondence and reprint requests to Geert E.C. Verhoeven, at the above address. E-mail: geertverhoeven@hotmail.com Received 4 April 2009; accepted for publication 24 February 2010. doi: 10.1111/j.1740-8261.2010.01693.x Experienced observers are aware of the importance of radiographic quality in the diagnosis of CHD and are familiar with quality criteria. Radiographic quality assessment and FCI-scoring is performed in the same reading session. If a radiograph is considered assessable, the observer automatically indicates that the radiograph contains all the information that is needed to make a correct evaluation of the hip status. Both interobserver agreement on assessability and on final scoring remain low despite improved subjective radiographic quality (assessable or not). 12,13 Therefore, the National Committee for Inherited Skeletal Disorders in Belgium (NCISD) introduced a detailed and morphometric-based technical quality assessment to standardize the evaluation of radiographic quality and thus to improve interobserver agreement. This technical quality assessment is done in a separate reading session. We are not aware of the existence of a technical quality assessment of hip-extended radiographs. Our aim was to describe the effect of a technical quality assessment on interobserver agreement in the diagnosis of CHD. Materials and Methods Thirty observers, recruited from the universities of Bern, Ghent, Giessen, Utrecht, Zurich, and from private practices, who are members of the Flanders Orthopaedic Working Group, took part in the study as described previously. 12,13 First, observers determined whether they would accept the radiograph for evaluation (radiograph is assessable or not). Secondly, observers indicated whether 498

Vol.51,No. 5 EFFECT OF A TECHNICAL QUALITYASSESSMENT 499 s Canon EOS 300D, Tokyo, Japan. wdigimizer s, MedCalc Software bvba, Mariakerke, Belgium. Table 1. Fédération Cynologique Internationale (FCI) Protocol as Presented at the FCI-Workshop on Hip Dysplasia in Copenhagen on March 18, 2006 Procedure for HD screening: the following rules (that focus on positioning and radiographic quality) are to be followed: 1. The minimum size of the X-ray film must be such as to totally include the pelvis and if possible both knees. 2. The technical quality of the radiographs has to be such as to allow an accurate screening procedure of the hip joint. The margo acetabularis must be clearly visualized. The positioning of the dog must ensure that the pelvis is symmetrical and not rotated along neither the long nor the short axis. Both ossa femoris must be parallel to each other and to the sagittal plane. The knees must be pronated so that the patellae are projected in sulcus intercondylaris on femur and held in a position close to the table. The patellae should be in contact with the line through the top of the fabellae. each coxofemoral joint was dysplastic or not, and third provided a final score for each joint according to FCIregulations (A, B, C, D, or E), 15 irrespective of whether the radiograph had been deemed assessable or not. Observers were classified as either experienced (9) or inexperienced (21). The NCISD database was searched for dogs having more than one radiograph submitted where there were obvious positional differences. The selection was done irrespective of the official FCI-score. However, dogs with obvious degenerative joint disease or coxofemoral luxation were excluded. Furthermore, dogs with radiographic signs of malformation of the pelvis or lumbosacral joint were not included. Fifty radiographs from 25 dogs were selected. Observers evaluated the radiographs individually, unaware of the fact that for every dog two radiographs were present. The radiographic quality on all 50 radiographs was assessed in a separate reading session according to the NCSID scoring system (technical quality assessment). The technical quality assessment was routinely implemented in the NCISD evaluation procedure for radiographs of clientowned dogs since August 2002. The authors are not aware of a similar system being used in other countries. The technical quality assessment is a method to quantify the FCI, OFA, and BVA/KC guidelines to reach the perfect position of the dog. The cut-off values of the technical quality assessment have been obtained empirically. The radiographic quality on all 50 radiographs was assessed by one individual (F.C.) in a separate reading session according to the NCSID scoring system (technical quality assessment). For the technical quality assessment, all radiographs were digitized using a digital camera and stored in a computer. Linear measurements that represent the metrical position of the iliac wings (width at comparable regions), femurs (deviation from the body axis), and patellae (positioning related to the fabellae), compared with the ideal position as described by the FCI-regulations, 15 which are still used today (Table 1), were performed on these digitized images with a software program.w Although Digimizer s can be used for nonlinear measurements, comparison of surfaces of the pelvic cavities is extremely time consuming and therefore not applicable in routine screening practice. Pelvic symmetry along the body axis was measured by determining the ratio of the width of both iliac wings (ratio ¼ 1 if 100% symmetry). Additionally, pelvic symmetry was judged by the visual symmetry of the obturator foramina, acetabula, and the femoral head coverage by comparing each pelvic half, determined by a line drawn through the body long axis. Pelvic rotation along the transverse axis was not judged. Parallelism of both femora was determined by measuring the angle of deviation between the long axis of each femur with the long axis of the body. By drawing a line through the femoral long axis and a line through the top of both fabellae, the position of the patella was evaluated: if both lines divided the patella in half, the femur was considered sufficiently extended (Fig. 1). Fig. 1. A, pelvic symmetry by metrical width of the iliac wings. B, the long axis of the body. C, femurs positioned parallel to the long axis of the spine. D, patellas symmetrically centered over the femurs and intersecting a line connecting the proximal margins of corresponding fabellae.

500 VERHOEVEN ET AL. 2010 Subscores were created on the original radiographs for the contrast/exposure of the image, as determined by the visibility (from invisible to complete visibility) of the dorsal acetabular edge, physis of the femoral head, subchondral acetabular bone, and trabecular bone structure. Additional subscores were introduced for determining motion blur and artifacts of the film. In an excel worksheet (Fig. 2), the results of the technical control were automatically processed to obtain a final score for the technical quality of the radiograph. In this system, a radiograph is considered as technically acceptable if the overall score is at least 60%, and if the subscores for symmetry of the pelvis, positioning of the femurs and patellae, motion blur and File number 4590/193 Technical evaluation 1. Symmetry of the Pelvis: 20 pts. score: 17 1.1 Symmetry of the iliac wings: 5 pts. score: 5 1.2 Symmetry of the pelvic cavity: 5 pts. score: 4 1.3 Symmetry of the foramina obturatoria: 5 pts. score: 4 1.4 Symmetry of the acetabuli-fhco: 5 pts. score: 4 2. Parallellism of the femora : 20 pts. score: 18 2.1 Di-convergens of the right femora: 10 pts. score: 8 2.2 Di-conv ergens of the left femora: 10 pts. score: 10 3. Positioning of the patella: 20 pts. score: 14 3.1 Latero-medial position of the right patella: 5 pts. score: 4 3.2 Latero-medial position of the left patella: 5 pts. score: 4 3.3 Proximo-distal position of the right patella: 5 pts. score: 3 3.4 Proximo-distal position of the left patella: 5 pts. score: 3 4. Exposure and contrast: 30 pts. score: 29 4.1 Visibility of the right dorsal acetabular edge: 5 pts. score: 5 4.2 Visibility of the left dorsal acetabular edge: 5 pts. score: 5 4.3 Visibility of the right epifysar growthline: 5 pts. score: 5 4.4 Visibility of the left epifysar growthline: 5 pts. score: 5 4.5 Visibility of the right trabecular bonestructure: 5 pts. score: 4 4.6 Visibility of the left trabecular bonestructure: 5 pts. score: 5 5. Motion blur: 5 pts. score: 5 6. Cleaning: 5 pts. score: 5 subcores % 1. 85 2. 90 3. 70 4. 97 5. 100 6. clean Score (%): 88 Very good Fig. 2. Detailed technical quality assessment of the radiograph as presented in Fig. 1. unclean film are at least 50%. Additionally, the film is only accepted if the subscores for contrast and exposure of the film are at least 60%. This implies that a radiograph will be rejected for final scoring by the NCISD committee if the overall quality is o60%. Additionally, rejection will occur in the following instances: inadequate pelvic symmetry (left right ratio 41.12), a difference of 451 between the long axis of the femurs and the body axis, touching of the patella with the cortex of the femur, the patella being above or below the line through the top of both fabellae and in dogs where less than half of the structure such as the dorsal acetabular edge, trabecular bone structure, physis of the femoral head, or subchondral acetabular bone is visible. The assessability of a radiograph was defined as the percentage of observers that scored the radiograph as assessable. Interobserver agreement was calculated as the percentage of observers that made the same diagnosis: dysplasia or no dysplasia. An agreement score for FCI was derived in the same way. All evaluations on 50 radiographs by 30 observers were used in the statistical analysis. It was tested whether the radiographic technical quality had an effect on the interobserver agreement score (dysplasia/nondysplasia and FCI-scores) by the mixed model with dog as random effect and the radiographic quality as continuous fixed effect. Additionally, it was tested whether there was an agreement between the subjective quality control (assessability) as done by the observers and the radiographic technical quality assessment using the generalized linear model with as response variable the binary variable technical quality assessment and as covariate the assessability of the radiographs. Furthermore, the experienced group was compared with the inexperienced group in respect to the agreement between assessability and technical quality assessment by the generalized linear mixed model with as response variable the binary variable agreement assessability technical quality assessment (0 ¼ different, 1 ¼ same), as covariate experience (experienced vs. inexperienced) and finally radiograph as random effect. Level of significance was 0.001. Results Interobserver agreement for experienced observers on assessability was low (68%). 13 Seventeen radiographs of 14 dogs passed the technical quality assessment, which would imply that for 11 dogs, new radiographs would be requested. The average total score given on the accepted radiographs is 78.4% (range 70.5 84.2%) and for unacceptable radiographs 57.3% (range 32.6 73.7%). Radiographs with an overall score of 460% can still be rejected because of reasons mentioned above, which explains the overlap of unacceptable or acceptable radiographs. The description of the final scoring performed by the experi-

Vol.51,No. 5 EFFECT OF A TECHNICAL QUALITYASSESSMENT 501 Table 2. Evaluations of Experienced Observers on Technically Accepted Radiographs Experienced observers % Gr:A % Gr:B % Gr:C % Gr:D % Gr:E Sum A þ B % Dyspl Sum C þ D þ E 1 24 18 35 24 0 41 59 59 2 6 29 24 41 0 35 65 65 3 24 24 41 12 0 47 53 53 4 0 29 53 18 0 29 100 71 5 6 35 47 12 0 41 59 59 6 18 41 29 12 0 59 41 41 7 0 24 47 24 6 24 82 77 8 0 47 29 24 0 47 53 53 9 18 24 29 29 0 41 59 59 % GrA E, % grades A E; sum A þ B, sum of % A and % B score; % dyspl, % dysplastic hips; sum C þ D þ E, sum of % C, % D, and % E. enced group on the technically approved radiographs is listed in Tables 2 and 3. The effect of technical quality assessment is narrowing of range of final scoring in dog 1 and narrowing of range of final scoring and polarization toward dysplasia in dogs 3 and 7, but not significantly. After technical quality assessment, the range of scoring decreases by 1 or 2 grades in seven dogs for experienced observers and in six dogs for inexperienced observers, but not significantly. Unfortunately, overall interobserver agreement in determining dysplasia/nondysplasia and FCI-scoring did not increase significantly with the increasing radiographic quality, irrespective of observer experience (Po0.0001). There was a significant relationship between technical quality assessment and assessability (Po0.0005), with the odds ratio of obtaining a positive technical quality assessment equal to 1.1067 (95% CI: 1.046 1.172) for each percentage increase in assessability, which implies that increased subjective quality of the radiograph correlates with increased technical quality. This relationship between technical quality assessment and assessability in the inexperienced group was 70%, whereas in the experienced group it was 62%. This difference is not significant (P ¼ 0.173). Table 3. Results According to Experienced Observers (nine) on the Technically Accepted Radiographs Dog % Dyspl % Gr A þ B % Gr C 1 89 11 78 3 100 0 89 6 22 78 22 7 100 0 44 8 89 11 44 12 33 89 11 13 11 100 0 14 100 0 0 16 78 22 78 17 100 0 100 18 44 67 33 21 100 0 44 22 11 100 0 23 22 89 11 % Dyspl, % dysplastic hips; % Gr A þ B, sum of % grade A and B; %GrC,%gradeC. Figure 3 is an example of two radiographs of the same dog (dog 1). During technical quality control, radiograph 3A was rejected because the femurs were not parallel. Nevertheless, five experienced observers accepted radiograph 3A and scored the dog A C, while eight experienced observers accepted radiograph 3B and scored the same dog in the range of B D. The most important trait for rejection of a radiograph after technical quality assessment is nonparallelism of the femurs, followed by the lack of exposure/ contrast (Table 4). Discussion Although technical quality assessment was able to decrease the range of final scoring in seven dogs and six dogs (experienced and inexperienced observers, respectively), overall interobserver agreement on dysplasia/nondysplasia and final scoring does not significantly increase with increasing radiographic quality using a technical quality control. Surprisingly, there was inconsistency between the final score and the diagnosis of dysplasia/nondysplasia in some observers (Table 3). This could be explained if some observers considered some grade B dogs as dysplastic, despite the FCI-regulations that define B as nondysplastic. High quality radiographs decrease the range of final scoring and causes a polarization toward dysplasia/nondysplasia in some dogs. Nevertheless, even with adequate radiographic quality, the range in final scoring remains unacceptably high, because the same dog can receive a score that permits unrestrictive breeding or a score that would exclude this dog from the breeding population. Only dogs 3 and 7 are considered undoubtedly dysplastic after technical quality assessment in the experienced group. Introducing a technical quality assessment in a separate reading session is therefore not sufficient to improve interobserver agreement. The range of scoring is affected more by the experts individual scoring method than by film quality. There is a significant relationship between the assessability of the radiograph and the evaluation after a technical quality assessment. Surprisingly, the relationship between assessability and technical quality assessment is less in the experienced group than in the inexperienced

502 VERHOEVEN ET AL. 2010 Fig. 3. Radiographs of dog 1. (A) Was rejected after technical quality assessment because of nonparallelism of femurs while (B) was accepted. Five of nine experienced observers accepted (A) and eight accepted (B). Range of scores among the nine experienced observers was A D for this dog. Table 4. Reason for Rejecting Radiographs Using the Technical Quality Assessment Trait NRR Total score 0 Symmetry of the pelvis (1) 0 Parallelism of the femora (2) 9 Positioning of the patellae (3) 0 Contrast and exposure time (4) 4 2and4only 2 Total and 2 only 2 Total and 3 only 2 Total, 1, and 2 only 3 Total, 1, and 3 only 1 Total, 1, and 4 only 1 Total, 2, and 3 only 1 Total, 2, and 4 only 1 Total, 3, and 4 only 1 Total,1,2,and3only 1 Total,1,2,and4only 1 Total,2,3,and4only 1 Total,1, 2, 3, and 4 3 Total of technically refused radiographs 33 Total of evaluated radiographs 50 NRR, number of rejected radiographs per trait or cluster of traits. group. A possible explanation may be that experienced observers automatically combine the quality assessment of the radiograph with scoring in a single step, whereas inexperienced observers tend to separate the quality assessment and scoring. Experienced observers may feel more confident in their image interpretation. Both radiographs of dog 1 (Fig. 3) are examples of how a flaw might affect final scoring. The nonparallel femora might obscure laxity and therefore incorrect upgrading might occur (Fig. 3A). The parallel femora and correct positioning of the patellae have been obtained by rotation of the pelvis over its transverse axis (Fig. 3B). In practice, radiograph 3B will be accepted for both assessability and technical quality assessment and scored. Which evaluation for this dog (B, C, or D) is correct remains obscure because a gold standard to evaluate CHD radiographically on a hip-extended radiograph does not exist. The NCISD technical quality assessment-method is not as objective as should be. Nevertheless, 60 points are entirely based on (25) or helped by (35) linear measurements or marks on the digitized radiographs. Additional nonlinear measurements can be implemented, if necessary. Visual control of exposure and contrast will always remain subjective. If one is able to measure the Norberg angle or femoral head coverage, one can assume that exposure and contrast is at least acceptable. Despite the drawbacks of this technical quality assessment, this quality assessment is a more structured approach to perform quality assessment of a radiograph. Because the cut-off values in the technical quality assessment system have not been validated, the

Vol.51,No. 5 EFFECT OF A TECHNICAL QUALITYASSESSMENT 503 system needs to be investigated further. The technical quality assessment is designed to reach the most perfect radiographic quality, which is difficult, if not impossible, to obtain. Therefore, radiographs with a quality value of more than 60% after technical quality assessment were regarded as acceptable for reading. There is some bias by eliminating obviously osteoarthritic and subluxated dogs from the data, leading to lower interobserver agreement. However, in reality, radiographs of obviously affected dogs are often withheld from official screening. The results of this study apply only for dogs without anatomic malformations of the pelvis or lumbosacral joint, rendering correct position impossible. Interobserver agreement on these dogs was not investigated. In conclusion, interobserver agreement in diagnosing CHD and in providing final scoring does not significantly improve with improved image quality, as determined by a technical quality assessment. Seemingly, increasing radiographic quality alone is insufficient in improving interobserver agreement to an acceptable level. This may indicate that the FCI-classification system for scoring hips is inadequately defined or is used in a subjective manner according the experts individual scoring method, which leads to different evaluations. 1. Cardinet GH, Lust G. The international symposium on hip dysplasia and osteoarthritis in dogs. J Am Vet Med Assoc 1997;210:1417 1418. 2. Powers MY, Biery DN, Lawler DF, et al. Use of the caudolateral curvilinear osteophytes as an early marker for future development of osteoarthritis associated with hip dysplasia in dogs. J Am Vet Med Assoc 2004;225:233 237. 3. Van Der Velden NA. Hip dysplasia in dogs. Vet Q 1983;5:3 10. 4. Bouw J. Hip dysplasia and dog breeding. Vet Q 1982;4:173 181. 5. Smith GK, Biery DN, Rhodes WH. Between- and within-radiologist accuracy of subjective hip scoring of the ventrodorsal hip-extended radiograph (abstract), In Proceedings, International Symposium Hip Dysplasia Osteoarthritis Dogs, 1996, 20. 6. Stur I, Ko ppel E, Schro der K. Populationsgenetische Aspekte der Hu ftgelenksdysplasie (HD)-Diagnostik beim Hund-Bewertung unter Berücksichtigung differierender HD-Befunde. Wien Tierärztl Mschr 1996;83:91 97. 7. Leppa nen M, Mäki K, Juga J. Factors affecting hip dysplasia in German shepherd dogs in Finland: efficacy of the current improvement programme. J Small Anim Pract 2000;41:19 23. 8. Paster ER, Lafond E, Biery DN, et al. Estimates of prevalence of hip dysplasia in golden retrievers and Rottweilers and the influence of bias on published prevalence figures. J Am Vet Med Assoc 2005;226: 387 392. 9. Smith GK. Advances in diagnosing canine hip dysplasia. J Am Vet Med Assoc 1997;210:1451 1457. 10. Kapatkin AS, Fordyce HH, Mathew PD, Smith GK. Canine hip dysplasia: the disease and its diagnosis. Compend Contin Educ Vet 2002;24:526 535. 11. Coopman F, Paepe D, Van Bree H, Saunders JH. 2004 The prevalence and evolution of canine hip dysplasia in Belgium (abstract), In Annual Scientific Conference Proceedings, The European Association of Veterinary REFERENCES Diagnostic Imaging and The European College of Veterinary Diagnostic Imaging, Ghent, Belgium, 68 12. Verhoeven G, Coopman F, Duchateau L, Saunders J, Van Ryssen B, Van Bree H. Inter-observer agreement in the diagnosis of canine hip dysplasia using the standard ventrodorsal hip-extended radiographic method. J Small Anim Pract 2007;48:387 393. 13. Verhoeven G, Coopman F, Duchateau L, Bosmans T, Van Ryssen B, Van Bree H. Interobserver agreement on the assessability of standard ventrodorsal hip extended radiographs and its effect on agreement in the diagnosis of canine hip dysplasia and on routine FCI-scoring. Vet Radiol Ultras 2009;50:259 263. 14. Popovitch CA, Smith GK, Gregor TP, Shofer FS. Comparison of susceptibility for hip dysplasia between Rottweilers and German shepherd dogs. J Am Vet Med Assoc 1995;206:648 650. 15. Brass W. Hip dysplasia in dogs. J Small Anim Pract 1989;30:166 170. 16. Corley EA, Keller GG, Lattimer JC, Ellersieck MR. Prevalence of canine hip dysplasia obtained from the standard ventrodorsal radiographic projection. J Am Vet Med Assoc 1997;211:1142 1146. 17. Gibbs C. The BVA/KC scoring scheme for control of hip dysplasia: interpretation of criteria. Vet Rec 1997;13:275 284. 18. Morgan JP, Wind A, Davidson AT. Hip dysplasia in hereditary bone and joint diseases in the dogs. Hannover: Schlu tersche, 2000;131 171. 19. Keller G. The use of health databases and selective breeding: a guide for dog and cat breeders and owners, 4th ed. Columbia, MO: Orthopedic Foundation of Animals, 2003;16 21. 20. Coopman F, Comhaire F, Schoonjans F, de Brabander K. Hip dysplasia research at Ghent University; towards a new approach to assess hip quality? Abstract and poster presented at the international congress on advances in canine and feline genomics. Davis, CA, USA: University of California, 2006.