Validation of a Computer-Assisted Method for Measurement of Radiographic Wear in Total Hip Arthroplasty Using All Polyethylene Cemented Acetabular Components Jean Langlois,1 Amine Zaoui,1 Caroline Scemama,1 John Martell,2 Charles Bragdon,3 Moussa Hamadouche1 1
Service de chirurgie orthope´dique et traumatologique, Hoˆpital Cochin, APHP, Universite´ Rene´ Descartes, 27 rue du Faubourg Saint-Jacques 75679, Paris CEDEX 14, France, 2Department of Orthopaedics, University of Chicago, Chicago, Illinois, 3Harris Orthopaedic Laboratory and the Adult Reconstructive Unit of the Department of Orthopaedic Surgery of Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts 02114 Received 23 September 2014; accepted 26 October 2014 Published online 6 January 2015 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jor.22777
ABSTRACT: Although cemented all polyethylene (PE) cups have been routinely used in total hip arthroplasty for decades, no computer-assisted method for measurement of radiographic wear has ever been specifically validated for these implants. Using a validated hip phantom model, AP plain hip radiographs were obtained consecutively for eight simulated wear positions. A version of Martell’s Hip Analysis Suite software dedicated to all polyethylene sockets was used by three different examiners of varied experience. Bias (mean, standard deviation and 95% confidence interval limit), repeatability (standard deviation and 95% limit) and reproducibility (standard deviation and 95% limit) for two-dimensional wear measurements were assessed, as recommended by the current ASTM guidelines. Using this protocol, the dedicated software showed an overall mean bias of 0.089 0.060 mm (mean SD), and 0.118 mm for 95% CI limit. Repeatability (intra examiner) standard deviation and 95% limit were respectively 0.106 mm and 0.292 mm. Reproducibility (inter examiner) standard deviation and 95% limit were respectively 0.112 mm and 0.308 mm. Martell Hip Analysis for all PE cemented cups is a reliable and low-cost instrument in the assessment of wear, despite being less precise than its original version dedicated to cementless components. ß 2015 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 33:417–420, 2015. Keywords: total hip replacement; polyethylene wear; error measurement; precision measurement; accuracy measurement
The number of total hip arthroplasty (THA) surgical procedures in the United States is expected to reach 572,000 per year by the year 2030.1 As this procedure is getting more and more reliable, an increasing number of younger patients are now THA candidates. However it has been shown that this population might present accelerated polyethylene wear. To solve this problem, many refinements in polyethylene (PE) processing have been proposed including cross-linking with or without the addition of an anti-oxidant such as vitamin E.2 These have showed excellent results in hip simulator studies and in clinical practice, leading to a continued interest in using PE in THA. In addition, there is increasing recent evidence that cemented fixation of acetabular components might be more reliable than that of cementless components beyond the first postoperative decade.3,4 This phenomenon could be related to accelerated PE wear in uncemented compared to cemented acetabular components.5 As polyethylene wear remains a major factor in failure of THA through particle-induced osteolysis and resultant aseptic loosening,6 new challenges for in vivo measurement of wear in THA implants are a matter of major concern. Indeed, wear measurement is critical to adequately evaluate new materials and can also help to decide when to monitor patients more closely and when to consider revision surgery. Though various
Conflict of interest: None. Grant sponsor: Fulbright and the Arthur Sachs Foundations. Correspondence to: Jean Langlois (Tel: þ00 (33) 1 58 41 30 82; F: 00 (33) 1 44 89 78 22; E-mail: [email protected]) # 2015 Orthopaedic Research Society. Published by Wiley Periodicals, Inc.
radiographic methods have been developed to measure polyethylene wear, including manual or computerassisted plain radiography, two- or three-dimensional techniques, or radiostereometric analysis,7,8 to the best of our knowledge, none has ever been specifically dedicated to all PE cemented implants. The purposes of this study were to assess the reliability of a specific version of Martell’s Hip Analysis Suite, a computerized semiautomatic edge detection method, dedicated to measure wear of all PE cemented implants.
METHODS Phantom Setup, Radiographs and Displacement Protocol The development of the physical phantom and the protocol used for calculating accuracy and precision has been previously described in detail.9 In the current study, the model consisted of a 52-mm outer diameter all PE cemented acetabular component (Durasul1, Zimmer, Warsaw, IN) fixed in a hemi-pelvis bone model (Sawbones, Pacific Research Labs, Vashon, WA) at 45˚ of abduction and 15˚ of anteversion, and a femoral component (Exafit1, Zimmer) with a 22.2 mm chromium-cobalt femoral head. Three separate verniers, all at right angles to each other, with 1 mm resolution were used to control independently the motion of the femoral head in three planes. To simulate PE wear, the component was previously worn so that the femoral head was free to be moved in various defined positions within the PE cup. The radiographic settings used for these projections were 76 kV and 100 mA and anteroposterior radiographic projections of the hip phantom were obtained simulating the pubic symphysis centered in the radiographic film. The stem was sequentially moved in two coronal directions, medially and superiorly, to simulate 0, 50, 100, 150, and 200 mm of femoral head migration. Anteroposterior X-rays were taken JOURNAL OF ORTHOPAEDIC RESEARCH MARCH 2015
417
418
LANGLOIS ET AL.
in each position. Two-dimensional vector magnitudes were calculated for each displacement interval using the known superior and medial motion. We thereby were able to simulate eight standard positions of PE wear, extending from 0.05 to 0.28 mm. Method of Radiological Wear Measurement All radiographic images were captured as DICOM images and subsequently converted and saved in an eight bit TIFF format using a plug-in for Photoshop called Dicom Access (DesAcc, Inc Chicago, IL) The Martell Hip Analysis Suite software (version 8.0.1.4.3; UCTech the University of Chicago) is a validated computer-assisted semi-automated edge detection method. A two-dimensional wear analysis was performed in the coronal plane. The cup center was determined from a circle drawn at the interface between the socket and the cement to delimit the acetabular component (Fig. 1). A specific all-polyethylene socket registry was added by the developer to this dedicated version of the Martell system, in order to enhance grey level contrast at the bone cement/PE interface. All radiographic measurements were performed by three independent examiners, previously familiarized with the software, who were blinded to the extent of wear. Evaluations were based on fivefold measurements of each femoral head position performed in a random order. The investigators were: (i) an expert; (ii) a well-trained; and (iii) an inexperienced examiner. Reliability Assessment A number of high quality publications9–12 have employed different definitions with various meanings for the evaluation of accuracy and precision over the past decade. However, in order to be consistent with recently updated ASTM recommendations,13 we calculated three parameters: bias, repeatability and reproducibility, defined as follows. Please
note that according to these latest ASTM standards, accuracy, defined as the “closeness of agreement between a test result and an accepted reference value,” involving a combination of random errors (imprecision) and systematic errors (bias), is presently described as a “confusing word that should be abandoned.” Precision, “the closeness of agreement between paired test results obtained under stipulated conditions,” has also been described as an inaccurate term that should be replaced either by repeatability (experimental conditions, where the test method is conducted by a single operator using one set of equipment in a short period of time) or reproducibility (experimental conditions, where the test method is conducted in several laboratories, or by different operators, apparati or environmental conditions). Consequently for this study, we first defined a set of 40 bias values (systematic errors) for each examiner; a bias was calculated as the difference between a test result and its accepted reference value. Three parameters were computed based on the set of biases for each examiner: (i) average; (ii) standard deviation; and (iii) 95% limit (equivalent to the standard deviation multiplied by 1.96) of the set of bias. In addition, we paired measurements within each examiner’s data (n ¼ 16 pairs were created per each examiner) and computed their repeatability parameters, sr, and r respectively, standard deviation and 95% limit from the distribution of differences between paired measures. Finally, we paired measurements per couple of examiners (40 pairs were created per each couple of examiners) and computed their reproducibility parameters, sR, and R, respectively standard deviation and 95% limit from the distribution of differences between those paired measures. For historical comparison purposes, we also computed our results using previously published methods for calculating accuracy and precision. Precision was therefore presented as suggested in previous ASTM177–10 standards,14 equivalent
Figure 1. Martell’s Hip Analysis Suite “all-polyethylene socket dedicated” sample AP view. Note the radiographic beam is not perfectly aligned over the symphysis pubis, but a few millimeters closer to the hip. Under magnification is underlined the dedicated tool that has been added by the developer in order to enhance grey level contrast at the bone cement/PE interface. JOURNAL OF ORTHOPAEDIC RESEARCH MARCH 2015
419
WEAR MEASUREMENT FOR CEMENTED HIP COMPONENTS
Table 1. Martell Software for Cemented Cups Reliability Assessment Using the Latest ASTM Recommendations Bias
Examiner 1 Examiner 2 Examiner 3 Overall
Mean
SD
0.076 0.063 0.127 0.089
0.047 0.049 0.084 0.060
Repeatability 95% limit 0.092 0.096 0.165 0.118
sr
Reproducibility
r
0.051 0.117 0.150 0.106
sR
0.141 0.321 0.415 0.292
Ex 1 vs 2 Ex 1 vs 3 Ex 2 vs 3 Overall
0.112 0.105 0.118 0.112
R 0.310 0.291 0.324 0.308
Values are expressed in millimeters. SD, set of bias standard deviation; sr, repeatability standard deviation; r, repeatability 95% limit; sR, reproducibility standard deviation; R, reproducibility 95% limit.
to 1.96 multiplied by the standard deviation of a set of paired measurements, and accuracy as a root mean square error considering a constant bias.12 Statistical Analysis Statistical analyses were performed using Real Statistics Resource Pack software, version 2.10 (Miller Place, NY). A p-value
Service de chirurgie orthope´dique et traumatologique, Hoˆpital Cochin, APHP, Universite´ Rene´ Descartes, 27 rue du Faubourg Saint-Jacques 75679, Paris CEDEX 14, France, 2Department of Orthopaedics, University of Chicago, Chicago, Illinois, 3Harris Orthopaedic Laboratory and the Adult Reconstructive Unit of the Department of Orthopaedic Surgery of Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts 02114 Received 23 September 2014; accepted 26 October 2014 Published online 6 January 2015 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jor.22777
ABSTRACT: Although cemented all polyethylene (PE) cups have been routinely used in total hip arthroplasty for decades, no computer-assisted method for measurement of radiographic wear has ever been specifically validated for these implants. Using a validated hip phantom model, AP plain hip radiographs were obtained consecutively for eight simulated wear positions. A version of Martell’s Hip Analysis Suite software dedicated to all polyethylene sockets was used by three different examiners of varied experience. Bias (mean, standard deviation and 95% confidence interval limit), repeatability (standard deviation and 95% limit) and reproducibility (standard deviation and 95% limit) for two-dimensional wear measurements were assessed, as recommended by the current ASTM guidelines. Using this protocol, the dedicated software showed an overall mean bias of 0.089 0.060 mm (mean SD), and 0.118 mm for 95% CI limit. Repeatability (intra examiner) standard deviation and 95% limit were respectively 0.106 mm and 0.292 mm. Reproducibility (inter examiner) standard deviation and 95% limit were respectively 0.112 mm and 0.308 mm. Martell Hip Analysis for all PE cemented cups is a reliable and low-cost instrument in the assessment of wear, despite being less precise than its original version dedicated to cementless components. ß 2015 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 33:417–420, 2015. Keywords: total hip replacement; polyethylene wear; error measurement; precision measurement; accuracy measurement
The number of total hip arthroplasty (THA) surgical procedures in the United States is expected to reach 572,000 per year by the year 2030.1 As this procedure is getting more and more reliable, an increasing number of younger patients are now THA candidates. However it has been shown that this population might present accelerated polyethylene wear. To solve this problem, many refinements in polyethylene (PE) processing have been proposed including cross-linking with or without the addition of an anti-oxidant such as vitamin E.2 These have showed excellent results in hip simulator studies and in clinical practice, leading to a continued interest in using PE in THA. In addition, there is increasing recent evidence that cemented fixation of acetabular components might be more reliable than that of cementless components beyond the first postoperative decade.3,4 This phenomenon could be related to accelerated PE wear in uncemented compared to cemented acetabular components.5 As polyethylene wear remains a major factor in failure of THA through particle-induced osteolysis and resultant aseptic loosening,6 new challenges for in vivo measurement of wear in THA implants are a matter of major concern. Indeed, wear measurement is critical to adequately evaluate new materials and can also help to decide when to monitor patients more closely and when to consider revision surgery. Though various
Conflict of interest: None. Grant sponsor: Fulbright and the Arthur Sachs Foundations. Correspondence to: Jean Langlois (Tel: þ00 (33) 1 58 41 30 82; F: 00 (33) 1 44 89 78 22; E-mail: [email protected]) # 2015 Orthopaedic Research Society. Published by Wiley Periodicals, Inc.
radiographic methods have been developed to measure polyethylene wear, including manual or computerassisted plain radiography, two- or three-dimensional techniques, or radiostereometric analysis,7,8 to the best of our knowledge, none has ever been specifically dedicated to all PE cemented implants. The purposes of this study were to assess the reliability of a specific version of Martell’s Hip Analysis Suite, a computerized semiautomatic edge detection method, dedicated to measure wear of all PE cemented implants.
METHODS Phantom Setup, Radiographs and Displacement Protocol The development of the physical phantom and the protocol used for calculating accuracy and precision has been previously described in detail.9 In the current study, the model consisted of a 52-mm outer diameter all PE cemented acetabular component (Durasul1, Zimmer, Warsaw, IN) fixed in a hemi-pelvis bone model (Sawbones, Pacific Research Labs, Vashon, WA) at 45˚ of abduction and 15˚ of anteversion, and a femoral component (Exafit1, Zimmer) with a 22.2 mm chromium-cobalt femoral head. Three separate verniers, all at right angles to each other, with 1 mm resolution were used to control independently the motion of the femoral head in three planes. To simulate PE wear, the component was previously worn so that the femoral head was free to be moved in various defined positions within the PE cup. The radiographic settings used for these projections were 76 kV and 100 mA and anteroposterior radiographic projections of the hip phantom were obtained simulating the pubic symphysis centered in the radiographic film. The stem was sequentially moved in two coronal directions, medially and superiorly, to simulate 0, 50, 100, 150, and 200 mm of femoral head migration. Anteroposterior X-rays were taken JOURNAL OF ORTHOPAEDIC RESEARCH MARCH 2015
417
418
LANGLOIS ET AL.
in each position. Two-dimensional vector magnitudes were calculated for each displacement interval using the known superior and medial motion. We thereby were able to simulate eight standard positions of PE wear, extending from 0.05 to 0.28 mm. Method of Radiological Wear Measurement All radiographic images were captured as DICOM images and subsequently converted and saved in an eight bit TIFF format using a plug-in for Photoshop called Dicom Access (DesAcc, Inc Chicago, IL) The Martell Hip Analysis Suite software (version 8.0.1.4.3; UCTech the University of Chicago) is a validated computer-assisted semi-automated edge detection method. A two-dimensional wear analysis was performed in the coronal plane. The cup center was determined from a circle drawn at the interface between the socket and the cement to delimit the acetabular component (Fig. 1). A specific all-polyethylene socket registry was added by the developer to this dedicated version of the Martell system, in order to enhance grey level contrast at the bone cement/PE interface. All radiographic measurements were performed by three independent examiners, previously familiarized with the software, who were blinded to the extent of wear. Evaluations were based on fivefold measurements of each femoral head position performed in a random order. The investigators were: (i) an expert; (ii) a well-trained; and (iii) an inexperienced examiner. Reliability Assessment A number of high quality publications9–12 have employed different definitions with various meanings for the evaluation of accuracy and precision over the past decade. However, in order to be consistent with recently updated ASTM recommendations,13 we calculated three parameters: bias, repeatability and reproducibility, defined as follows. Please
note that according to these latest ASTM standards, accuracy, defined as the “closeness of agreement between a test result and an accepted reference value,” involving a combination of random errors (imprecision) and systematic errors (bias), is presently described as a “confusing word that should be abandoned.” Precision, “the closeness of agreement between paired test results obtained under stipulated conditions,” has also been described as an inaccurate term that should be replaced either by repeatability (experimental conditions, where the test method is conducted by a single operator using one set of equipment in a short period of time) or reproducibility (experimental conditions, where the test method is conducted in several laboratories, or by different operators, apparati or environmental conditions). Consequently for this study, we first defined a set of 40 bias values (systematic errors) for each examiner; a bias was calculated as the difference between a test result and its accepted reference value. Three parameters were computed based on the set of biases for each examiner: (i) average; (ii) standard deviation; and (iii) 95% limit (equivalent to the standard deviation multiplied by 1.96) of the set of bias. In addition, we paired measurements within each examiner’s data (n ¼ 16 pairs were created per each examiner) and computed their repeatability parameters, sr, and r respectively, standard deviation and 95% limit from the distribution of differences between paired measures. Finally, we paired measurements per couple of examiners (40 pairs were created per each couple of examiners) and computed their reproducibility parameters, sR, and R, respectively standard deviation and 95% limit from the distribution of differences between those paired measures. For historical comparison purposes, we also computed our results using previously published methods for calculating accuracy and precision. Precision was therefore presented as suggested in previous ASTM177–10 standards,14 equivalent
Figure 1. Martell’s Hip Analysis Suite “all-polyethylene socket dedicated” sample AP view. Note the radiographic beam is not perfectly aligned over the symphysis pubis, but a few millimeters closer to the hip. Under magnification is underlined the dedicated tool that has been added by the developer in order to enhance grey level contrast at the bone cement/PE interface. JOURNAL OF ORTHOPAEDIC RESEARCH MARCH 2015
419
WEAR MEASUREMENT FOR CEMENTED HIP COMPONENTS
Table 1. Martell Software for Cemented Cups Reliability Assessment Using the Latest ASTM Recommendations Bias
Examiner 1 Examiner 2 Examiner 3 Overall
Mean
SD
0.076 0.063 0.127 0.089
0.047 0.049 0.084 0.060
Repeatability 95% limit 0.092 0.096 0.165 0.118
sr
Reproducibility
r
0.051 0.117 0.150 0.106
sR
0.141 0.321 0.415 0.292
Ex 1 vs 2 Ex 1 vs 3 Ex 2 vs 3 Overall
0.112 0.105 0.118 0.112
R 0.310 0.291 0.324 0.308
Values are expressed in millimeters. SD, set of bias standard deviation; sr, repeatability standard deviation; r, repeatability 95% limit; sR, reproducibility standard deviation; R, reproducibility 95% limit.
to 1.96 multiplied by the standard deviation of a set of paired measurements, and accuracy as a root mean square error considering a constant bias.12 Statistical Analysis Statistical analyses were performed using Real Statistics Resource Pack software, version 2.10 (Miller Place, NY). A p-value
