Published online before print September 27, 2006, 10.1148/radiol.2412050960
(Radiology 2006;241:386-391.)
© RSNA, 2006
Interobserver Agreement for Letournel Acetabular Fracture Classification with Multidetector CT: Are Standard Judet Radiographs Necessary?1
Kenjirou Ohashi, MD,
Georges Y. El-Khoury, MD,
Khalil W. Abu-Zahra, MD and
Kevin S. Berbaum, PhD
1 From the Department of Radiology, University of Iowa Roy J. and Lucille A. Carver College of Medicine, 200 Hawkins Dr, Iowa City, IA 52242. From the 2004 RSNA Annual Meeting. Received June 8, 2005; revision requested August 2; revision received October 10; accepted November 14; final version accepted January 2, 2006.
Address correspondence to K.O. (e-mail: kenjirou-ohashi{at}uiowa.edu).
 |
ABSTRACT
|
|---|
Purpose: To retrospectively evaluate interobserver agreement for Letournel acetabular fracture classification with radiography alone and multidetector computed tomography (CT) alone and to retrospectively assess whether standard Judet views lead to a change in the classification.
Materials and Methods: Institutional review board approval was obtained; informed consent was not required for this HIPAA-compliant study, which included 101 imaging studies performed in 99 patients (78 male, 21 female; mean age, 43 years; age range, 15–86 years) with acetabular fractures. Two musculoskeletal radiologists independently classified the fractures with radiography alone and multidetector CT alone. Multiplanar reformatted and three-dimensional (3D) CT images were reviewed at a computer workstation. Readers were shown radiographs at the end of multidetector CT image reading to see if this would change the multidetector CT–based classification. 
Values were calculated to assess interobserver agreement. For surgically treated patients, the McNemar test was used to compare the accuracy of readers' classifications. The reference standard was a combination of preoperative radiographic and multidetector CT image findings and intraoperative findings.
Results: Interobserver agreement was moderate ( = 0.42) with radiography and substantial ( = 0.70) with multidetector CT. Multidetector CT classification was changed in two cases (one case for each reader) after standard Judet views were added. In 73 surgically treated patients, agreement with the surgeons' classification was higher with multidetector CT than with radiography (P < .01 for one reader, P = .06 for the other reader).
Conclusion: There is substantial interobserver agreement for Letournel acetabular fracture classification with multiplanar reformatted and 3D multidetector CT images. Standard Judet pelvic radiographs add little information for changing the multidetector CT classification.
© RSNA, 2006
|
INTRODUCTION
|
|---|
Judet et al (1) defined the anterior and posterior columns of the acetabulum and introduced a classification system and surgical approaches in 1964. The Letournel classification system is now widely used by orthopedic surgeons to understand fracture patterns, plan treatment, and evaluate outcomes. In 2003, Beaulé et al (2) assessed interobserver agreement for Letournel classification by using randomly selected cases that had been classified at the time of surgery. To our knowledge, no previous study has been performed to evaluate interobserver agreement for Letournel classification by using consecutive patients with acetabular fractures. Beaulé et al (2) evaluated the role of transverse computed tomography (CT) in Letournel classification, but they did not evaluate the role of CT apart from radiography.
Currently, CT is routinely performed to aid preoperative planning in patients with acetabular fractures (3–6). In addition, advances in multidetector CT technology and high-speed computing have made it possible to review multiplanar reformatted images in any arbitrary plane and obtain high-quality three-dimensional (3D) images for viewing in any direction. However, since Letournel classification is based on standard pelvic radiography consisting of anteroposterior and both oblique (Judet) views (7), both Judet views are routinely requested in patients suspected of having acetabular fractures.
Thus, the purpose of this study was to retrospectively evaluate interobserver agreement for Letournel acetabular fracture classification with radiography alone and multidetector CT alone and to retrospectively assess whether standard Judet views lead to a change in the classification.
|
MATERIALS AND METHODS
|
|---|
Patients
Our institutional review board approved this retrospective study and patient informed consent was not required. This study was conducted in accordance with the federal Health Insurance Portability and Accountability Act. We included consecutive patients with acute acetabular fractures who had undergone both pelvic CT and standard pelvic radiography at our institution. (At our institution, all patients who have or are suspected of having acetabular fractures undergo multidetector CT of the pelvis.) During a 38-month period (August 2001 through September 2004), 101 acetabular fractures in 99 patients (78 male, 21 female; mean age, 43 years; age range, 15–86 years) were available for review. Seventeen patients had been excluded because a suboptimal radiographic technique was used. In addition, 14 acetabular fractures had been excluded because they were not depicted by radiography or CT. There were 44 right and 57 left acetabular fractures. Two patients had bilateral fractures. An author (K.W.A.) reviewed medical records for clinical data, including mechanism of injury, interval between injury and radiographic study, number of radiographic views, interval between radiographic and CT studies, and postsurgical classification of acetabular fractures in surgically treated patients. The interval between injury and surgery was also recorded for surgically treated patients.
Imaging Techniques
Radiographic studies of the pelvis were performed with computed radiography (Fuji CR900; Fuji Medical Systems, Stamford, Ct). Radiographs in at least three views, including the anteroposterior and both oblique (Judet) views with or without inlet and outlet views, were obtained.
CT studies were performed with either a four–detector row (Aquilion; Toshiba American Medical Systems, Tustin, Calif) or a six–detector row (Emotion 6; Siemens, Malvern, Pa) helical CT scanner. With the four–detector row CT scanner, scanning parameters were 120–135 kVp, 175–525 mAs (0.5–1.5-second scanning time per gantry rotation), 320–400-mm field of view, 512 x 512 matrix, 2.0–3.0-mm collimation, 3.5–4.5-mm table travel per rotation, 2.0–3.0-mm reconstruction thickness, 66%–75% overlap (1.5–2.0 mm), and standard soft-tissue kernel. With the six–detector row CT scanner, scanning parameters were 130 kVp, 120–250 mAs (0.5-second scanning time per gantry rotation), 320–500-mm field of view, 512 x 512 matrix, 2.0-mm collimation, 18-mm table travel per rotation, 2.5-mm reconstruction thickness, 50% overlap (1.2 mm), and medium-smooth (soft-tissue) kernel.
Classification of Acetabular Fractures
Two musculoskeletal radiologists (K.O., G.Y.E., 10 and 13 years of experience with Letournel classification of acetabular fractures, respectively) retrospectively and independently classified acetabular fractures according to Letournel classification by using radiography alone and multidetector CT alone. Radiograph reading was followed by multidetector CT image reading at least 4 weeks later; images were viewed in a different order to avoid recall bias. The readers used percentages to indicate the likelihood of each acetabular fracture type (eg, 80% transverse fracture vs 20% T-shaped fracture). No unclassified category was used, and a rating of 50% versus 50% was not allowed (one dominant type had to be chosen). After each multidetector CT image was read, readers were shown radiographs to determine if radiographs influence Letournel classification based on multidetector CT image findings. The influence of radiographs was recorded.
Image Review Methods
Radiographs were reviewed on a two- or four-monitor system in our intradepartmental picture archiving and communication system network (Kodak System 4; Eastman Kodak, Rochester, NY). A combination of multiplanar reformatted and volume-rendered 3D multidetector CT images were reviewed on a computer workstation (Vitrea 2, version 3.5; Vital Images, Plymouth, Minn). Two types of 3D images were used according to reader preference. One was a volume-rendered view with a lighting source applied to enhance surface detail. The other was a volume-rendered view with a variable degree of transparency applied to the bone density range (3D see through).
Interobserver Agreement and Disagreement
Number of agreements and types of disagreement between the readers were recorded (K.W.A.) for radiography alone and multidetector CT alone. The results were analyzed to see if there would be notable tendencies in agreement or disagreement among the acetabular fracture classifications.
Statistical Analysis
Interobserver agreement for acetabular fracture classification was tested by using statistical software (BMDP4F, release 8.0; BMDP Statistical Software, Crosse's Green, Cork, Ireland) to calculate values between two readers who used radiography alone and multidetector CT alone to classify 101 acetabular fractures. The fracture type that the radiologist indicated as most likely was used in the analysis. According to Landis and Koch (8), a value of less than 0.00 indicates poor agreement; a value of 0.00–0.20, slight agreement; a value of 0.21–0.40, fair agreement; a value of 0.41–0.60, moderate agreement; a value of 0.61–0.80, substantial agreement; and a value of 0.81–1.00, almost perfect agreement.
For surgically treated patients, the McNemar test (BMDP4F, release 8.0) was used to compare the agreement of radiographic classifications with postsurgical classifications (reference standard) to the agreement of multidetector CT classifications with postsurgical classifications. A postsurgical classification was the orthopedic surgeons' final classification, and it was based on preoperative radiographic and multidetector CT image findings and intraoperative findings. Again, the most likely fracture type was used in the analysis.
|
RESULTS
|
|---|
Clinical Data
Of the 99 patients, 65 (66%) were involved in a motor vehicle accident, 15 (15%) fell from some height, and 19 (19%) experienced some other type of injury. Seventy-eight patients (79%) underwent radiography on the day the injury occurred. The interval between the radiographic study and the onset of the injury ranged from 0 to 11 days. The average number of radiographic views obtained was 3.7 (range, three to five views; median, three views). Eighty-one patients (82%) underwent radiography and CT on the same day; the interval between radiography and CT ranged from 0 to 2 days.
Seventy-six patients (77%) underwent surgery; the mean interval between injury and surgery was 4.8 days (range, 0–26 days). Postsurgical classification of acetabular fractures was available for 73 patients (73 fractures), whereas three fractures were not classified according to Letournel classification. There were 18 (24%) posterior wall, three (4%) posterior column, no (0%) anterior wall, three (4%) anterior column, six (8%) transverse, eight (11%) T-shaped, seven (10%) posterior column and posterior wall, 15 (21%) transverse and posterior wall, eight (11%) both columns, and five (7%) anterior column and posterior hemitransverse acetabular fractures.
Interobserver Agreement
The values for interobserver agreement for classification of 101 acetabular fractures were 0.42 when only radiographs were viewed and 0.70 when only multidetector CT images were viewed (8).
Agreement in Surgically Treated Patients
Postsurgical classifications assigned by the surgeons were treated as the reference standard, and values were higher for classification with multidetector CT alone than for classification with radiography alone for both readers (reader 1, 0.73 vs 0.54; reader 2, 0.66 vs 0.52). We used a pair of McNemar tests (one for each reader) to compare the agreement of multidetector CT classifications and radiographic classifications with the surgeons' classifications (Table 1). For reader 1, the agreement of multidetector CT classifications with the surgeons' classifications was significantly higher than the agreement of radiographic classifications with the surgeons' classifications (P < .01). For reader 2, the agreement of multidetector CT classifications with the surgeons' classifications was higher than the agreement of radiographic classifications with the surgeons' classifications (P = .06).
Multidetector CT Classifications Supplemented with Additional Radiography
In two cases of acetabular fracture (one fracture for each reader), the multidetector CT classification was changed after readers viewed additional radiographs. One fracture first classified as 60% posterior wall fracture versus 40% transverse and posterior wall fracture after multidetector CT images were viewed was classified as a transverse and posterior wall fracture (100% certainty) when the radiographs were viewed after multidetector CT. A minimally displaced transverse fracture component was not initially recognized on the multidetector CT images. One fracture first classified as 70% anterior column and posterior hemitransverse fracture versus 30% T-shaped fracture after multidetector CT images were viewed was classified as a T-shaped fracture (100% certainty) when the radiographs were viewed after multidetector CT. A superior pubic ramus fracture that was not continuous with the acetabular fracture was thought to involve the anterior aspect of the acetabulum on the multidetector CT images. Even after these classifications were supplemented with additional radiographs, the observers' classifications were not consistent with the surgeons' classifications. Thus, additional radiography did not contribute to agreement with the surgeons' classifications.
Interobserver Agreement and Disagreement on Fracture Classification
There were apparent tendencies (Tables 2, 3) for anterior column fractures and closely related fractures or fractures with a common feature, such as a transverse fracture component (transverse, T-shaped, and transverse and posterior wall fractures), to be frequently associated with disagreements and for the rate of agreement to be higher for posterior wall fractures than for other types of fractures. However, the sample sizes for each fracture type were too small to allow statistical analysis. The number of agreements for each fracture type were compared between radiography and multidetector CT (Figure). The number of agreements was larger with multidetector CT than with radiography for all fracture types other than anterior wall fractures.
View larger version (31K):
[in this window]
[in a new window]
[Download PPT slide]
|
Graph shows number of agreements based on radiography and multidetector CT findings for 101 acetabular fractures. AC = anterior column fracture, AC PHT = anterior column (wall) and posterior hemitransverse fracture, AW = anterior wall fracture, Both Col = both columns fracture, PC = posterior column fracture, PC PW = posterior column and posterior wall fracture, PW = posterior wall fracture, Trans PW = transverse and posterior wall fracture.
|
|
|
DISCUSSION
|
|---|
We found that interobserver agreement for Letournel classification of acetabular fractures based on radiographic findings was moderate ( = 0.42), whereas Beaulé et al (2) found it to be substantial ( = 0.51–0.76). This may be related to the difference in sampling. We used a consecutive series of acetabular fractures that were seen on radiographs and CT images, whereas Beaulé et al used preclassified surgical cases. Beaulé et al also used nine guidelines for classification, which likely contributed to the higher interobserver agreement. As for the contribution of CT to fracture classification, Beaulé et al found no significant improvement in the values between radiographic classification with and without CT, but they used only transverse CT images and radiographs. Our results, however, indicate significant improvement in observer agreement with postsurgical classification for multidetector CT ( = 0.73 and 0.66) when compared with radiography ( = 0.54 and 0.52). Improved observer agreement was observed in each fracture type, with the exception of the anterior wall fracture type, which was rarely classified in our series (twice with radiography and once with multidetector CT). The advantages of viewing multiplanar reformatted and 3D CT images at a workstation may explain this difference.
Our results show that multiple radiographic views had little effect on the multidetector CT classification of acetabular fractures. On only two occasions did the readers change the multidetector CT classification after they viewed the radiographs. A minimally displaced transverse fracture component was not initially recognized at multidetector CT in one patient. In another patient, a superior pubic ramus fracture that was not continuous with the acetabular fracture appeared to involve the anterior aspect of the acetabulum on multidetector CT images. It has been shown that subtle fractures can be missed on 3D images (9). The anatomic boundary of the anterior column of the acetabulum may be less clear on cross-sectional multiplanar reformatted images than on 3D images (10). We believe these display modes should be used in a complementary fashion to maximize their benefits (11).
Obtaining images depicting both oblique views of the pelvis may not always be easy in patients with acetabular fractures. Even with adequate projections, image quality may be reduced by overlapping bowel gas and life-supporting tubes and lines. Scott et al (12) showed that perspective 3D images have advantages over multiple radiographic views because they provide unobstructed images of the pelvic bone. The advantages of avoiding multiple radiographic views have been discussed elsewhere (12). Fishman and Kuszyk (13) reported a change in care in 30% of patients with acetabular fractures when 3D images were obtained.
Certain types of fractures are diagnosed less consistently than others. Disagreements tend to occur more frequently in fracture types that have a common fracture component, such as a transverse fracture component (which is common to transverse, T-shaped, and transverse and posterior wall fractures). Harris et al (10) proposed redefining the anterior column to avoid ambiguity in diagnosis. Our results also showed a poor agreement associated with anterior column fractures. As noted in surgically treated patients, there are some fracture patterns that surgeons cannot clearly classify. It is likely that our observations support other authors' thoughts in that the Letournel classification is associated with intrinsic complexity and oversimplification (14,15). Harris et al (15) also proposed a new system for classification of acetabular fractures that would be based on CT findings.
Limitations of our study included a limited number of readers and a small sample size of certain types of fractures because of their rarity. We used surgeons' classifications in medical records as the reference standard for patients who underwent surgery; however, no reference standard was available for conservatively treated patients. We could not use the published guidelines for Letournel classification of acetabular fractures (2), as they were unavailable when we began this investigation. However, we used experienced readers who routinely classify acetabular fractures on the basis of Letournel classification in daily practice.
In summary, our results suggest interobserver agreement is moderate for readers who use Letournel acetabular fracture classification to classify radiographic findings; interobserver agreement significantly improves when multiplanar reformatted and 3D multidetector CT images are used. Standard Judet pelvic radiographs added little information for changing the acetabular fracture classification that was based on multidetector CT findings in surgically treated patients, with the surgeons' classification serving as the reference standard. This led us to conclude that Judet pelvic radiographs are not needed to classify acetabular fractures.
|
ADVANCES IN KNOWLEDGE
|
|---|
- Letournel acetabular fracture classification with radiography alone has moderate interobserver agreement ( = 0.42).
- Letournel acetabular fracture classification with multiplanar reformatted and three-dimensional images from multidetector CT data has substantial interobserver agreement ( = 0.70).
- In 73 (74%) surgically treated patients, agreement with the surgeons' classification was higher with multidetector CT than with radiography (P < .01 for one reader and P = .06 for the other reader).
- Standard Judet pelvic radiographs add little information for changing the multidetector CT classification of acetabular fractures.
|
FOOTNOTES
|
|---|
Abbreviations: 3D = three dimensional
Authors stated no financial relationship to disclose.
Author contributions: Guarantor of integrity of entire study, K.O.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, K.O., K.W.A.; clinical studies, K.O., G.Y.E., K.W.A.; statistical analysis, K.O., K.S.B.; and manuscript editing, K.O., G.Y.E., K.S.B.
|
References
|
|---|
- Judet R, Judet J, Letournel E. Fractures of the acetabulum: classification and surgical approaches for open reduction—preliminary report. J Bone Joint Surg Am 1964;46:1615–1646.[Free Full Text]
- Beaulé PE, Dorey FJ, Matta JM. Letournel classification for acetabular fractures: assessment of interobserver and intraobserver reliability. J Bone Joint Surg Am 2003;85-A:1704–1709.[Abstract/Free Full Text]
- Harley JD, Mack LA, Winquist RA. CT of acetabular fractures: comparison with conventional radiography. AJR Am J Roentgenol 1982;138:413–417.[Abstract/Free Full Text]
- Mack LA, Harley JD, Winquist RA. CT of acetabular fractures: analysis of fracture patterns. AJR Am J Roentgenol 1982;138:407–412.[Abstract/Free Full Text]
- Griffiths HJ, Standertskjöld-Nordenstam CG, Burke J, Lamont B, Kimmel J. Computed tomography in the management of acetabular fractures. Skeletal Radiol 1984;11:22–31.[CrossRef][Medline]
- Borrelli J Jr, Goldfarb C, Catalano L, Evanoff BA. Assessment of articular fragment displacement in acetabular fractures: a comparison of computerized tomography and plain radiographs. J Orthop Trauma 2002;16:449–456.[CrossRef][Medline]
- Letournel E. Fractures of the acetabulum. Heidelberg, Germany: Springer-Verlag, 1993.
- Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33:159–174.[CrossRef][Medline]
- Martinez CR, Di Pasquale TG, Helfet DL, Graham AW, Sanders RW, Ray LD. Evaluation of acetabular fractures with two- and three-dimensional CT. RadioGraphics 1992;12:227–242.[Abstract]
- Harris JH Jr, Lee JS, Coupe KJ, Trotscher T. Acetabular fractures revisited. I. Redefinition of the Letournel anterior column. AJR Am J Roentgenol 2004;182:1363–1366.
- Haveri M, Junila J, Suramo I, Lähde S. Multiplanar and 3D CT of acetabular fractures. Acta Radiol 1998;39:257–264.[Medline]
- Scott WW Jr, Fishman EK, Magid D. Acetabular fractures: optimal imaging. Radiology 1987;165:537–539.[Abstract/Free Full Text]
- Fishman EK, Kuszyk B. 3D imaging: musculoskeletal applications. Crit Rev Diagn Imaging 2001;42:59–100.[Medline]
- Brandser E, Marsh JL. Acetabular fractures: easier classification with a systematic approach. AJR Am J Roentgenol 1998;171:1217–1228.[Free Full Text]
- Harris JH Jr, Coupe KJ, Lee JS, Trotscher T. Acetabular fractures revisited. II. A new CT-based classification. AJR Am J Roentgenol 2004;182:1367–1375.
TOP
ABSTRACT
INTRODUCTION
