- Email: [email protected]
Lack of agreement between computer navigation and post-operative 2-dimensional computed tomography (CT) measurements for component and limb alignment in total knee arthroplasty (TKA)
THEKNE-02121; No of Pages 7 The Knee xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
The Knee
Lack of agreement between computer navigation and post-operative 2-dimensional computed tomography (CT) measurements for component and limb alignment in total knee arthroplasty (TKA) Ziad Dahabreh a, Corey J. Scholes a,⁎, Bruno Giuffre b, Myles R.J. Coolican a, David A. Parker a a b
Sydney Orthopaedic Research Institute, Chatswood, NSW, Australia Department of Radiology, Royal North Shore Hospital, NSW, Australia
a r t i c l e
i n f o
Article history: Received 21 May 2014 Received in revised form 1 July 2015 Accepted 15 July 2015 Available online xxxx Keywords: Computer-assisted surgery Computed tomography Alignment Mismatch TKA
a b s t r a c t Background: The purpose of this study was to assess the degree of mismatch between intraoperative navigation data using imageless computer navigation and post-operative CT scan measurements with respect to bone cuts, component and limb alignment during TKA. Methods: Intraoperative navigation data including bone cut verification and overall limb alignment during TKA was compared to postoperative CT measurements of component and limb alignment according to the Perth protocol. The proportion of cases with mismatch between navigation and CT measurements at two and three degree thresholds was identified. Results: In a total sample of 50 primary TKAs, 20% of cases showed a mismatch of more than two degrees between navigation and CT obtained measurements for coronal femoral alignment, 42% for femoral rotation, 16% for tibial component coronal alignment and 32% for overall limb alignment. Conclusion: Mismatch between intraoperative navigation data and postoperative CT measurements suggests that postoperative CT scan alignment data should be interpreted with caution. A surgeon should consider a multitude of factors when analysing component and limb alignment postoperatively. © 2015 Elsevier B.V. All rights reserved.
1. Introduction Optimal component positioning and subsequent limb alignment are important for long term knee function and implant longevity in total knee arthroplasty (TKA). Although the relationship between coronal malalignment and early failure in TKR remains controversial [1,2], the complications associated with poor component alignment [3], particularly in the rotation [4] and sagittal planes [5], cannot be ignored. The need for accurate TKA implant positioning has driven the development and implementation of real-time navigation with reduced incidence of alignment outliers [6,7]. However, reliable postoperative assessment of final alignment utilising plain radiography or two-dimensional CT (2D-CT) scanning remains challenging [8]. To date, no agreed gold standard imaging modality for assessing component and limb alignment following TKA has been established. Plain radiography is useful in identifying gross malposition but prone to error due to variations in knee flexion and leg rotation [9,10] and assessment of component rotation is not possible. Recently, attention has turned to computed tomography (CT) as a more accurate modality with
⁎ Corresponding author at: Suite 13, Level 1, 445 Victoria Avenue, Chatswood, NSW 2067, Australia. E-mail address: [email protected] (C.J. Scholes).
the benefit of assessing component rotation [11,12]. Although CT provides more detail and higher image quality than plain radiography, the lack of definitive landmarks limits interpretation of alignment data. Recently, the Perth CT based on two-dimensional (2D) CT has been described for assessment of TKA alignment and defines several alignment parameters [12]. The standardised and quantitative approach provides an objective, sensitive, numerical technique, which can undergo statistical analysis. However, routine use of CT in a clinical or research setting has limitations of poor inter- and intra-observer reliability, especially for component rotation [8,13]. Therefore, in situations where unsatisfactory outcomes of TKA are attributed to component malpositioning, surgeons should consider radiological measurement limitations. This has particular relevance to cases where measurements may inform the decision to perform revision arthroplasty or otherwise. Unfortunately, there is no strong evidence that a validation process had occurred before the orthopaedic community adopted the Perth CT protocol in clinical and research settings. Similarly, although computer navigation has undergone some validation in-vitro with respect to its motion capture qualities [14], such data has not been published for all systems available and there is a lack of criterion validation on knee alignment. Instead, validation of navigated TKA has been assessed with plain radiographs or CT [6,7], although the agreement between intraoperative component alignment derived from computer navigation and CT-based post-operative evaluation remains unknown.
http://dx.doi.org/10.1016/j.knee.2015.07.009 0968-0160/© 2015 Elsevier B.V. All rights reserved.
Please cite this article as: Dahabreh Z, et al, Lack of agreement between computer navigation and post-operative 2-dimensional computed tomography (CT) measurements for component..., Knee (2015), http://dx.doi.org/10.1016/j.knee.2015.07.009
2
Z. Dahabreh et al. / The Knee xxx (2015) xxx–xxx
Evaluation of TKA component positioning and limb alignment has been undertaken with a ‘group analysis’ [6], which determines the overall number of outliers in each group exceeding a pre-defined threshold (e.g. ± 3° from neutral alignment). Group analysis of 2D-CT data has demonstrated that computer navigated TKA reduced the incidence of outliers (±3°) from the neutral coronal mechanical alignment compared to conventional TKA [15]. However, this statistical method provides little guidance with respect to the discrepancy between intraoperative navigation and post-operative radiological measurements in an individual patient. Further research is required into the mismatch between intraoperative measurement techniques and post-operative image based measurements in individual patients. The purpose of this study therefore, was to determine the degree of mismatch between intraoperative navigation data and post-operative CT measurements with respect to bone cuts, as well as component and limb alignment following TKA. 2. Methods Between January 2010 and August 2012 two consultant orthopaedic surgeons performed a series of 50 primary TKAs in 41 patients to treat knee osteoarthritis. One consultant (MC) used the ORTHOSoft Navigation system (Zimmer Inc., Warsaw, USA) to implant Nexgen cruciateretaining (Zimmer Inc., Warsaw, USA) prostheses (cemented tibial component/uncemented porous coated femoral component), termed SN1. The other surgeon (DP) used the PrecisioN navigation system (Stryker Orthopaedics, Kalamazoo, USA) to implant all-cemented posterior stabilised Legion prostheses (Smith & Nephew Global, Memphis, USA), termed SN2. TKA was undertaken according to each surgeon's standard technique. Standard registration of landmarks and joint centres was conducted as previously described [16,17] with the lateral epicondyle defined as the most lateral prominence, whilst the medial epicondyle
was defined by its medial sulcus which allowed the software to generate the surgical trans-epicondylar axis (surTEA). Femoral and tibial bone cuts were verified in the coronal and sagittal planes (Fig. 1) and femoral rotation in the axial plane after the final bony cuts were made. Verification of the bone cuts was recorded before cementation of the implants. Final coronal limb alignment was recorded following implantation of definitive prostheses, including the polyethylene insert. A single CT scan of each patient's limb including the hip, knee and ankle joints was obtained between six weeks and three months postoperatively according to the Perth protocol [12]. Digital images were examined by an independent experienced consultant musculoskeletal radiologist, who was blinded to the intra-operative navigation measurements. The scanning protocol was standardised with the patient supine, the legs in neutral position, knees in full extension and the patellae pointing forwards. Based on the Perth CT protocol [12], coronal alignment of the femoral component was measured. The difference from 90° was recorded as the varus or valgus angle. Femoral component sagittal alignment was measured as the angle between a line from the centre of the femoral head to the centre of the femoral component and a line parallel to the posterior flange of the internal border of the femoral component. Femoral component rotation was measured on the axial images section at the level of epicondyles (identified by the radiologist) as the angle between a line passing through the epicondyles and a line parallel to the posterior aspect of the femoral condyles of the component (Fig. 2 — right). For coronal alignment of the tibial component, the difference from 90° was recorded as the varus or valgus angle. Tibial component sagittal alignment (slope) was measured as the angle between a line passing from the centre of the ankle mortise to centre of tibial component and a line perpendicular to the tibial base plate component. Coronal limb alignment was measured as the angle between a line from centre of femoral head to the centre of the femoral component and a line from the centre of the tibial component to the centre of the ankle
Fig. 1. Example of tibial (top) and femoral (bottom) bone cut validation produced by the imageless navigation system during total knee arthroplasty.
Please cite this article as: Dahabreh Z, et al, Lack of agreement between computer navigation and post-operative 2-dimensional computed tomography (CT) measurements for component..., Knee (2015), http://dx.doi.org/10.1016/j.knee.2015.07.009
Z. Dahabreh et al. / The Knee xxx (2015) xxx–xxx
3
Fig. 2. Example CT image for coronal alignment of the limb and prosthesis components (left) and femoral rotation (right) measurements using the Perth Protocol.
(Fig. 2 — left). All data was entered manually into a spreadsheet during measurement. Varus, component extension, and femoral external rotation were given a positive value. Valgus, component flexion and femoral internal rotation were given a negative value. The difference between CT and navigation measurements for each parameter was calculated for each individual. All continuous data was assessed for normality and equality of variance prior to statistical analysis. Patient age and BMI were compared between SN1 and SN2 using a Mann–Whitney independent samples t-test. The average discrepancy between navigation and CT was assessed whether it was significantly different from zero using Wilcoxon one-sample t-tests. Bland–Altman plots, Typical error [18] and Pearson's correlation were used to assess the agreement between modalities for measurement of femoral and tibial implant positioning. Categorical data such as side of surgery and gender were compared between groups using a Fisher's exact test. The incidence of discrepancies between navigation and CT that exceeded two degree and three degree thresholds (outlier incidence) was calculated for both groups and compared using Fisher's exact test. For alignment measures where no significant differences in incidence were detected, the outlier incidence was calculated for both groups combined and reported as a percentage. Alpha was set a-priori at five percent for all tests and all statistical analyses were performed in Minitab statistical software (version 16, Minitab Inc., MA, USA). 3. Results In a total sample of 50 primary TKAs, nine patients underwent simultaneous bilateral procedures. The 41 patients enrolled were elderly (authors to define if mean ± SD years), predominantly female (23 females) and overweight (BMI 30.1 ± 5.7). There were no significant differences between SN1 and SN2 for age at surgery, side of surgery or BMI. However, SN2 had a significantly (p b 0.05) higher proportion of females to males compared to SN1 (Table 1). Final post-implantation coronal limb alignment measurement with navigation was within three degrees of a neutral (0°) mechanical axis in 96% of cases, compared to 88% when measured by post-operative CT. The differences between navigation and CT measurements for femoral flexion (p b 0.01), femoral rotation (p = 0.013) and tibial slope (p b 0.01) were significantly different to zero (Table 2). Bland–Atlman plots revealed upper and lower limits of agreement ranging from −4.0 to 3.6° (femoral coronal) up to −4.8 to 7.0° (femoral rotation) (Fig. 3). Typical error for component measurements between modalities ranged from 1.4 to 2.1° (Table 2). The incidence of outliers outside two degrees and three degrees of
agreement between navigation and CT measurements ranged between 4.8% for tibial coronal alignment to 82.8% for tibial slope (Table 3). No differences in outlier incidence were observed between SN1 and SN2 for femoral coronal or rotational alignment, coronal tibial alignment or overall coronal limb alignment (Table 3). However, a significant difference (p b 0.01) in outlier incidence was observed between groups for sagittal alignment of the femoral component at the three degrees threshold and for tibial slope at both two degrees and three degrees thresholds (Table 3). Outlier incidence for the two groups combined at the three degrees threshold was between eight percent for tibial coronal alignment and 22% for femoral rotational alignment (Table 4). A similar pattern was observed at the two degrees threshold with a range of 16 to 42% (Table 4). Graphical analysis of the mismatch between navigation and CT for component alignment revealed considerable disagreement for femoral coronal alignment (Fig. 4 — A) and femoral rotation (Fig. 4 — B). Although tibial coronal alignment displayed the lowest incidence of outliers at the two degrees threshold (16%), the range of discrepancies between navigation and CT was also −6 to 6° (Fig. 4 — C). Mismatch was also observed between methods for final coronal alignment, with a higher incidence of outliers at both two degrees and three degrees thresholds than femoral and tibial coronal alignment alone (Fig. 4 — D), although the range was comparable to data for the individual components. Additional analysis of final coronal alignment outliers revealed that 69% of them occurred in TKAs that were within the two degrees threshold for the femoral and the tibial components, 19% in TKAs with outliers in the femoral values, six percent in TKAs with outliers in the tibial component, and six percent in outliers in both femoral and tibial values.
4. Discussion The purpose of this study was to determine the mismatch between intraoperative navigation data and post-operative CT measurements with respect to bone cuts, as well as component and limb alignment following TKA. Combined analysis of SN1 and SN2 data confirmed that Table 1 Patient demographics for S-N 1 and S-N 2.
N (knees) Mean age at surgery Side of surgery BMI (SD) Gender
SN 1
SN 2
p-Value
29 68.6 (50.2–84.1) Right = 17 Left = 12 29.4 (5.5) M = 17 F = 12
21 70.5 (61.1–80) R = 10 L = 11 31.1 (6.1) M=2 F = 19
– 0.38 – 0.35 0.004
Please cite this article as: Dahabreh Z, et al, Lack of agreement between computer navigation and post-operative 2-dimensional computed tomography (CT) measurements for component..., Knee (2015), http://dx.doi.org/10.1016/j.knee.2015.07.009
4
Z. Dahabreh et al. / The Knee xxx (2015) xxx–xxx
Table 2 Validated navigation and post-operative CT measurements (n = 50) of component and limb alignment. Mean difference is calculated by subtracting the navigation measurement from the CT measurement. Navigation (mean ± SD) (°) Femoral Coronal −0.2 (0.6) Flexion −2.7 (1.2) Rotation 1.2 (1.8) Tibial Coronal 0 (0.6) Slope 5.2 (2.0) Limb Alignment Coronal 0.4 (1.6)
CT (mean ± SD) (°)
Mean (±SD) difference (°)
One-sample t-test p-value
Typical error (95% CI) (°)
Pearson correlation
0 (1.8) −1.0 (2.0) 0.1 (1.8)
0.2 (1.9) 1.7 (2.1) −1.1 (3.0)
0.46 b0.01 0.013
1.4 (1.1–1.7) 1.5 (1.2–1.9) 2.1 (1.8–2.7)
0.01 0.20 −0.44
0.1 (1.9) 2.5 (2.5)
0.1 (1.9) −2.7 (2.9)
0.74 b0.01
1.4 (1.1–1.7) 2.1 (1.7–2.6)
0.05 0.17
0.7 (1.8)
0.3 (2.6)
satisfactory outlier incidence (± 3°) from neutral coronal limb alignment was achieved when assessed by both computer navigation (96%) and postoperative CT (88%); results that are comparable to previous reports [6,19]. However, intraoperative measurements did not match post-operative CT values in 8–42% of cases, depending on the component alignment measure and the threshold of acceptable deviation. Mismatch in overall limb alignment could not be attributed to a single component mismatch factor. The reported mismatch between modalities (computer navigation and 2D-CT) on a patient-by-patient basis has important implications for interpretation of alignment using 2D-CT following navigated TKA.
0.361
1.8 (1.6–2.2)
−0.1
Although CT imaging as a modality is accurate and detailed, points of reference that are utilised for further measurement and analysis of alignment are determined by experienced assessors, thus allowing for a certain level subjectivity. Further, the intra and inter-observer reliability of both modalities requires further study and should ideally be reported in subsequent work in this area. Indeed, the question remains regarding a gold standard to determine alignment, with promising results for 3-dimensional CT (3D-CT), showing significantly higher interobserver correlation than 2D-CT for femoral component rotation (0.91 vs 0.29), femoral coronal alignment (0.97 vs 0.65), and tibial component coronal alignment (0.89 vs 0.70) [8].
Fig. 3. Bland–Altman plots of agreement between navigation and CT measurements for lower limb coronal alignment, as well as component alignment in sagittal, coronal and rotation planes. Upper and lower limits are indicated with dashed lines above and below the mean.
Please cite this article as: Dahabreh Z, et al, Lack of agreement between computer navigation and post-operative 2-dimensional computed tomography (CT) measurements for component..., Knee (2015), http://dx.doi.org/10.1016/j.knee.2015.07.009
Z. Dahabreh et al. / The Knee xxx (2015) xxx–xxx Table 3 Outlier incidence at two and three degree thresholds for SN1 (n = 29) and SN2 (n = 21). 2° threshold S-N 1
Femoral Coronal (%) Flexion (%) Rotation (%) Tibial Coronal (%) Slope (%) Limb alignment Coronal (%)
3° threshold
S-N 2
S-N 1 p-Value (Fisher's exact test)
8 (27.6) 2 (9.5) 0.16 10 (34.5) 10 (47.6) 0.39 13 (44.8) 8 (38.1) 0.77
S-N 2
p-Value (Fisher's exact test)
4 (13.8) 2 (9.5) 1.0 3 (10.3) 8 (38.1) 0.04 6 (20.7) 5 (23.8) 1.0
7 (24.1) 24 (82.8)
1 (4.8) 0.12 6 (28.6) 0.0002
11 (37.9)
5 (23.8) 0.37
4 (13.8) 0 (0) 0.13 20 (69.0) 4 (19.1) 0.0006 7 (24.1) 2 (9.5)
0.27
Table 4 Mismatch between CT and computer navigation measurements (n = 50).
Femoral coronal Femoral rotation Tibia coronal Coronal limb alignment
Within 2° (%)
Out of 2° (%)
Within 3° (%)
Out of 3° (%)
40 (80) 29 (58) 42 (84) 34 (68)
10 (20) 21 (42) 8 (16) 16 (32)
44 (88) 39 (78) 46 (92) 41 (82)
6 (12) 11 (22) 4 (8) 9 (18)
Causes for discrepancies between navigation and CT measurements are multifactorial. Errors in navigation (e.g. landmark registration) and CT measurements co-vary and in some cases may be additive with respect to mismatch. Intra-operative femoral and tibial component alignments reflect the validated cut surface of the bone prior to cementation and implantation of prosthetic components. Therefore, mismatch between measurements for component alignment in the coronal and sagittal planes may be partly explained by changes in component alignment
5
following cementation and implantation, especially in the sagittal plane [20]. Points of reference in the femur and the tibia that define axes are also not standardised between navigation and CT. Navigation referencing is based on anatomical landmarks, whereas CT may be based on measured points of the tibial plateau diameters and visual landmarks in the distal femur (highest point of the notch) with the implants in-situ [12]. These factors will certainly have unequal net effects on discrepancies. Other factors may contribute to specific component measurement mismatch. For instance, CT measurement of femoral component sagittal alignment relies on referencing off the posterior condyle of the implant, thus potentially less accurate than coronal measurements of the component. Metal scatter on CT images may also reduce accuracy of measurements particularly in relation to the femoral component. Generally, several 2D-CT based points of reference are more prone to subjective bias than other well defined parameters. Sagittal distal femoral ‘highest point of the notch’ or epicondylar points to determine the transepicondylar axis may be less accurate than identifying the centre of the femoral head using digitised circles. The rotational position of the limb in the CT scanner could also influence the measured angles for component alignment and final limb alignment [21,22]. Tibial slope mismatch was high for both SN-1 and SN-2, with the rate of mismatch higher for SN-1. This may be attributed to the different desired intra-operative slope of the tibial cut which is higher in the SN-1 group, making post-operative CT measurement potentially less accurate if the limb is rotated inside the scanner. The difference in slope of the tibial component may also translate into mismatch in the coronal plane due to parallax error if the limb is rotated and/or flexed [21]. The high incidence of mismatch in femoral component rotation provides insight into potential causative factors, including intra-operative inaccuracy during registration of the epicondyles, or measurement error secondary to subjectively determining the epicondyles on CT. Firstly, the accuracy of determining the femoral rotation axis intraoperatively remains problematic, despite the use of navigation. That is, the surTEA is generated from the lateral and medial epicondyles that are manually identified by the surgeon, with undefined accuracy
Fig. 4. Mismatch between computer navigation and post-operative CT measurements for femoral coronal alignment (A), femoral rotation (B), tibial coronal alignment (C), and for final coronal alignment (D).
Please cite this article as: Dahabreh Z, et al, Lack of agreement between computer navigation and post-operative 2-dimensional computed tomography (CT) measurements for component..., Knee (2015), http://dx.doi.org/10.1016/j.knee.2015.07.009
6
Z. Dahabreh et al. / The Knee xxx (2015) xxx–xxx
[23–25]. Secondly, CT interpretation of femoral rotation is based on points and axes that are subject to inter- and intra-observer variability [26], which has important implications, as it is the basis for all applications of CT data in the clinical setting [27–29]. Variability in 2D-CT may be reduced by utilising 3D-CT imaging [8]. Future research may provide insight into the feasibility of standardising 3D-CT protocols for assessing component and limb alignment in TKA. Coronal tibial component alignment demonstrated the lowest number of outliers outside the two degree (16%) and three degree (eight percent) thresholds. This may be attributed to a combination of more reliably identifiable anatomical landmarks during tibial registration [25] as well as more clearly defined CT measurement points that are less prone to inter- or intra-observer variability [26]. This finding supports the speculation that the higher mismatch observed in other alignment parameters may be related to less reliable reference points. Final limb alignment should be independent of factors causing mismatch of component measurements as it uses reference points which are independent of individual component positioning. Errors in recorded navigation measurements of overall limb alignment may be related to varus or valgus stress that may have been applied passively and unintentionally, when the limb is held extended. Alternatively, rotational positioning of the limb in the CT scanner may lead to a misinterpretation of coronal alignment in a flexed and/or slightly rotated limb [21,22]. Nevertheless, significant discrepancies were found in overall limb alignment that had exceeded those for femoral or tibial component coronal alignment suggesting that mismatch between navigation and CT obtained final limb alignment figures is more likely to be the result of a true lack of correlation between the two modalities. Factors related to rotational limb positioning and of knee flexion may be responsible for the some of the discrepancies. The results of this study should be interpreted in light of some its limitations, which require exploration in future work. Firstly, postoperative CT measurements are based on the implanted prostheses. Small changes in component alignment (coronal and sagittal) might have occurred after preparation of bony surfaces, cementation and then impaction of the definitive prosthesis. A technique which allows validating component alignment after implantation may eliminate these differences and improve comparison. This may be possible with certain tibial component designs but will be much more technically demanding for the femoral component and almost impossible in the femoral sagittal plane. Secondly, the reliability of 2D-CT as well as for computer navigation in TKA is relatively ill-defined, thus comparison of measurements obtained with the two modalities remains challenging. Further, 2D-CT is the most widely available tool used by most surgeons for post-operative assessment of alignment and reflects common practice. However, it may be less accurate for verifying component and limb alignment than 3D-CT especially when evaluating component rotation [8]. Nevertheless, the introduction of 3D-CT is still limited by image scatter from the prosthetic components. To-date, the utility of 3D radiographs has been demonstrated for native segment alignment, but not for prosthetic components [30,31]. Lastly, the use of two separate combinations of surgeons, navigation systems and TKA implants may appear as a limitation, given the differences in tibial slope between implants. However, final intraoperative bone cut verification and alignment measurements should be independent of surgical technique, as would be subsequent CT analysis. 5. Conclusion This study shows that significant mismatch exists between intraoperative navigation and postoperative CT obtained measurements for component and limb alignment following TKA. Mismatch was highest for femoral rotation and lowest for tibial component coronal alignment. A surgeon should consider a multitude of factors when analysing component and limb alignment postoperatively. This is especially true when surgical decision making is based on CT results for rotational alignment of the femoral component or even for overall coronal limb
alignment. A method to perfectly determine component alignment for TKA remains elusive. Additional factors also complicate comparison and replication of this work, such as the surgeon, the geometry of arthroplasty components, and the specific navigation system. Future work should consider controlling as many of these factors before comparing between studies. Conflict of interest Ziad Dahabreh FRCS (Tr & Orth). No conflict of interest. Corey J. Scholes PhD. No conflict of interest. Bruno Giuffre DRANZCR. No conflict of interest. Myles Coolican FRACS. Institutional and Research support from Zimmer. David A. Parker FRACS. Institutional and Research support from Stryker and Smith and Nephew. Acknowledgements The authors wish to acknowledge the assistance of Amy Brierley and Joe Lynch at the Sydney Orthopaedic Research Institute and the staff of North Shore Radiology and Nuclear Medicine with the patient recruitment and data collection. This study was supported financially by the Sydney Orthopaedic Research Institute in-kind support, by Zimmer Inc. and Smith & Nephew. References [1] Kim Y-H, Park J-W, Kim J-S, Park S-D. The relationship between the survival of total knee arthroplasty and postoperative coronal, sagittal and rotational alignment of knee prosthesis. Int Orthop 2014;38:379–85. [2] Parratte S, Pagnano MW, Trousdale RT, Berry DJ. Effect of postoperative mechanical axis alignment on the fifteen-year survival of modern, cemented total knee replacements. J Bone Joint Surg Am 2010;92:2143–9. [3] Gallo J, Goodman SB, Konttinen YT, Wimmer MA, Holinka M. Osteolysis around total knee arthroplasty: a review of pathogenetic mechanisms. Acta Biomater 2013;9: 8046–58. [4] Berger RA, Crossett LS, Jacobs JJ, Rubash HE. Malrotation causing patellofemoral complications after total knee arthroplasty. Clin Orthop Relat Res 1998:144–53. [5] Lustig S, Scholes CJ, Stegeman TJ, Oussedik S, Coolican MR, Parker DA. Sagittal placement of the femoral component in total knee arthroplasty predicts knee flexion contracture at one-year follow-up. Int Orthop 2012;36:1835–9. [6] Hetaimish BM, Khan MM, Simunovic N, Al-Harbi HH, Bhandari M, Zalzal PK. Metaanalysis of navigation vs conventional total knee arthroplasty. J Arthroplasty 2012; 27:1177–82. [7] Zamora LA, Humphreys KJ, Watt AM, Forel D, Cameron AL. Systematic review of computer-navigated total knee arthroplasty. ANZ J Surg 2013;83:22–30. [8] Hirschmann MT, Konala P, Amsler F, Iranpour F, Friederich NF, Cobb JP. The position and orientation of total knee replacement components: a comparison of conventional radiographs, transverse 2D-CT slices and 3D-CT reconstruction. J Bone Joint Surg 2011;93:629–33. [9] Kannan A, Hawdon G, McMahon SJ. Effect of flexion and rotation on measures of coronal alignment after TKA. J Knee Surg 2012;25:407–10. [10] Skytta ET, Lohman M, Tallroth K, Remes V. Comparison of standard anteroposterior knee and hip-to-ankle radiographs in determining the lower limb and implant alignment after total knee arthroplasty. Scand J Surg 2009;98:250–3. [11] Blakeney WG, Khan RJ, Wall SJ. Computer-assisted techniques versus conventional guides for component alignment in total knee arthroplasty: a randomized controlled trial. J Bone Joint Surg Am 2011;93:1377–84. [12] Chauhan SK, Clark GW, Lloyd S, Scott RG, Breidahl W, Sikorski JM. Computer-assisted total knee replacement. A controlled cadaver study using a multi-parameter quantitative CT assessment of alignment (the Perth CT Protocol). J Bone Joint Surg 2004; 86:818–23. [13] Konigsberg B, Hess R, Hartman C, Smith L, Garvin K. Inter- and Intraobserver Reliability of Two-dimensional CT Scan for Total Knee Arthroplasty Component Malrotation. Clin Orthop Relat Res® 2014;472(1):212–7. [14] Lustig S, Fleury C, Goy D, Neyret P, Donell ST. The accuracy of acquisition of an imageless computer-assisted system and its implication for knee arthroplasty. Knee 2011;18:15–20. [15] Zhang GQ, Chen JY, Chai W, Liu M, Wang Y. Comparison between computerassisted-navigation and conventional total knee arthroplasties in patients
Please cite this article as: Dahabreh Z, et al, Lack of agreement between computer navigation and post-operative 2-dimensional computed tomography (CT) measurements for component..., Knee (2015), http://dx.doi.org/10.1016/j.knee.2015.07.009
Z. Dahabreh et al. / The Knee xxx (2015) xxx–xxx
[16]
[17]
[18] [19]
[20]
[21] [22] [23]
undergoing simultaneous bilateral procedures: a randomized clinical trial. J Bone Joint Surg Am 2011;93:1190–6. Doro LC, Hughes RE, Miller JD, Schultz KF, Hallstrom B, Urquhart AG. The reproducibility of a kinematically-derived axis of the knee versus digitized anatomical landmarks using a knee navigation system. Open Biomed Eng J 2008;2:52–6. Widmer BJ, Scholes CJ, Lustig S, Conrad L, Oussedik SI, Parker DA. Intraoperative computer navigation parameters are poor predictors of function 1 year after total knee arthroplasty. J Arthroplasty 2013;28:56–61. Hopkins WG. Measures of reliability in sports medicine and science. Sports Med 2000;30:1–15. Cheng T, Zhao S, Peng X, Zhang X. Does computer-assisted surgery improve postoperative leg alignment and implant positioning following total knee arthroplasty? A meta-analysis of randomized controlled trials? Knee Surg Sports Traumatol Arthrosc 2012;20:1307–22. Catani F, Biasca N, Ensini A, Leardini A, Bianchi L, Digennaro V, et al. Alignment deviation between bone resection and final implant positioning in computernavigated total knee arthroplasty. J Bone Joint Surg Am 2008;90:765–71. Lonner JH, Laird MT, Stuchin SA. Effect of rotation and knee flexion on radiographic alignment in total knee arthroplasties. Clin Orthop Relat Res 1996:102–6. Radtke K, Becher C, Noll Y, Ostermeier S. Effect of limb rotation on radiographic alignment in total knee arthroplasties. Arch Orthop Trauma Surg 2010;130:451–7. Robinson M, Eckhoff DG, Reinig KD, Bagur MM, Bach JM. Variability of landmark identification in total knee arthroplasty. Clin Orthop Relat Res 2006;442:57–62.
7
[24] Yau WP, Chiu KY, Tang WM. How precise is the determination of rotational alignment of the femoral prosthesis in total knee arthroplasty: an in vivo study. J Arthroplasty 2007;22:1042–8. [25] Yau WP, Leung A, Liu KG, Yan CH, Wong LL, Chiu KY. Interobserver and intraobserver errors in obtaining visually selected anatomical landmarks during registration process in non-image-based navigation-assisted total knee arthroplasty. J Arthroplasty 2007;22:1150–61. [26] Victor J, Van Doninck D, Labey L, Innocenti B, Parizel PM, Bellemans J. How precise can bony landmarks be determined on a CT scan of the knee? Knee 2009;16:358–65. [27] Amiot LP, Poulin F. Computed tomography-based navigation for hip, knee, and spine surgery. Clin Orthop Relat Res 2004:77–86. [28] Jazrawi LM, Birdzell L, Kummer FJ, Di Cesare PE. The accuracy of computed tomography for determining femoral and tibial total knee arthroplasty component rotation. J Arthroplasty 2000;15:761–6. [29] Nishihara S, Sugano N, Ikai M, Sasama T, Tamura Y, Tamura S, et al. Accuracy evaluation of a shape-based registration method for a computer navigation system for total knee arthroplasty. J Knee Surg 2003;16:98–105. [30] Folinais D, Thelen P, Delin C, Radier C, Catonne Y, Lazennec JY. Measuring femoral and rotational alignment: EOS system versus computed tomography. Orthop Traumatol Surg Res 2013;99:509–16. [31] Guenoun B, Zadegan F, Aim F, Hannouche D, Nizard R. Reliability of a new method for lower-extremity measurements based on stereoradiographic three-dimensional reconstruction. Orthop Traumatol Surg Res 2012;98:506–13.
Please cite this article as: Dahabreh Z, et al, Lack of agreement between computer navigation and post-operative 2-dimensional computed tomography (CT) measurements for component..., Knee (2015), http://dx.doi.org/10.1016/j.knee.2015.07.009
Contents lists available at ScienceDirect
The Knee
Lack of agreement between computer navigation and post-operative 2-dimensional computed tomography (CT) measurements for component and limb alignment in total knee arthroplasty (TKA) Ziad Dahabreh a, Corey J. Scholes a,⁎, Bruno Giuffre b, Myles R.J. Coolican a, David A. Parker a a b
Sydney Orthopaedic Research Institute, Chatswood, NSW, Australia Department of Radiology, Royal North Shore Hospital, NSW, Australia
a r t i c l e
i n f o
Article history: Received 21 May 2014 Received in revised form 1 July 2015 Accepted 15 July 2015 Available online xxxx Keywords: Computer-assisted surgery Computed tomography Alignment Mismatch TKA
a b s t r a c t Background: The purpose of this study was to assess the degree of mismatch between intraoperative navigation data using imageless computer navigation and post-operative CT scan measurements with respect to bone cuts, component and limb alignment during TKA. Methods: Intraoperative navigation data including bone cut verification and overall limb alignment during TKA was compared to postoperative CT measurements of component and limb alignment according to the Perth protocol. The proportion of cases with mismatch between navigation and CT measurements at two and three degree thresholds was identified. Results: In a total sample of 50 primary TKAs, 20% of cases showed a mismatch of more than two degrees between navigation and CT obtained measurements for coronal femoral alignment, 42% for femoral rotation, 16% for tibial component coronal alignment and 32% for overall limb alignment. Conclusion: Mismatch between intraoperative navigation data and postoperative CT measurements suggests that postoperative CT scan alignment data should be interpreted with caution. A surgeon should consider a multitude of factors when analysing component and limb alignment postoperatively. © 2015 Elsevier B.V. All rights reserved.
1. Introduction Optimal component positioning and subsequent limb alignment are important for long term knee function and implant longevity in total knee arthroplasty (TKA). Although the relationship between coronal malalignment and early failure in TKR remains controversial [1,2], the complications associated with poor component alignment [3], particularly in the rotation [4] and sagittal planes [5], cannot be ignored. The need for accurate TKA implant positioning has driven the development and implementation of real-time navigation with reduced incidence of alignment outliers [6,7]. However, reliable postoperative assessment of final alignment utilising plain radiography or two-dimensional CT (2D-CT) scanning remains challenging [8]. To date, no agreed gold standard imaging modality for assessing component and limb alignment following TKA has been established. Plain radiography is useful in identifying gross malposition but prone to error due to variations in knee flexion and leg rotation [9,10] and assessment of component rotation is not possible. Recently, attention has turned to computed tomography (CT) as a more accurate modality with
⁎ Corresponding author at: Suite 13, Level 1, 445 Victoria Avenue, Chatswood, NSW 2067, Australia. E-mail address: [email protected] (C.J. Scholes).
the benefit of assessing component rotation [11,12]. Although CT provides more detail and higher image quality than plain radiography, the lack of definitive landmarks limits interpretation of alignment data. Recently, the Perth CT based on two-dimensional (2D) CT has been described for assessment of TKA alignment and defines several alignment parameters [12]. The standardised and quantitative approach provides an objective, sensitive, numerical technique, which can undergo statistical analysis. However, routine use of CT in a clinical or research setting has limitations of poor inter- and intra-observer reliability, especially for component rotation [8,13]. Therefore, in situations where unsatisfactory outcomes of TKA are attributed to component malpositioning, surgeons should consider radiological measurement limitations. This has particular relevance to cases where measurements may inform the decision to perform revision arthroplasty or otherwise. Unfortunately, there is no strong evidence that a validation process had occurred before the orthopaedic community adopted the Perth CT protocol in clinical and research settings. Similarly, although computer navigation has undergone some validation in-vitro with respect to its motion capture qualities [14], such data has not been published for all systems available and there is a lack of criterion validation on knee alignment. Instead, validation of navigated TKA has been assessed with plain radiographs or CT [6,7], although the agreement between intraoperative component alignment derived from computer navigation and CT-based post-operative evaluation remains unknown.
http://dx.doi.org/10.1016/j.knee.2015.07.009 0968-0160/© 2015 Elsevier B.V. All rights reserved.
Please cite this article as: Dahabreh Z, et al, Lack of agreement between computer navigation and post-operative 2-dimensional computed tomography (CT) measurements for component..., Knee (2015), http://dx.doi.org/10.1016/j.knee.2015.07.009
2
Z. Dahabreh et al. / The Knee xxx (2015) xxx–xxx
Evaluation of TKA component positioning and limb alignment has been undertaken with a ‘group analysis’ [6], which determines the overall number of outliers in each group exceeding a pre-defined threshold (e.g. ± 3° from neutral alignment). Group analysis of 2D-CT data has demonstrated that computer navigated TKA reduced the incidence of outliers (±3°) from the neutral coronal mechanical alignment compared to conventional TKA [15]. However, this statistical method provides little guidance with respect to the discrepancy between intraoperative navigation and post-operative radiological measurements in an individual patient. Further research is required into the mismatch between intraoperative measurement techniques and post-operative image based measurements in individual patients. The purpose of this study therefore, was to determine the degree of mismatch between intraoperative navigation data and post-operative CT measurements with respect to bone cuts, as well as component and limb alignment following TKA. 2. Methods Between January 2010 and August 2012 two consultant orthopaedic surgeons performed a series of 50 primary TKAs in 41 patients to treat knee osteoarthritis. One consultant (MC) used the ORTHOSoft Navigation system (Zimmer Inc., Warsaw, USA) to implant Nexgen cruciateretaining (Zimmer Inc., Warsaw, USA) prostheses (cemented tibial component/uncemented porous coated femoral component), termed SN1. The other surgeon (DP) used the PrecisioN navigation system (Stryker Orthopaedics, Kalamazoo, USA) to implant all-cemented posterior stabilised Legion prostheses (Smith & Nephew Global, Memphis, USA), termed SN2. TKA was undertaken according to each surgeon's standard technique. Standard registration of landmarks and joint centres was conducted as previously described [16,17] with the lateral epicondyle defined as the most lateral prominence, whilst the medial epicondyle
was defined by its medial sulcus which allowed the software to generate the surgical trans-epicondylar axis (surTEA). Femoral and tibial bone cuts were verified in the coronal and sagittal planes (Fig. 1) and femoral rotation in the axial plane after the final bony cuts were made. Verification of the bone cuts was recorded before cementation of the implants. Final coronal limb alignment was recorded following implantation of definitive prostheses, including the polyethylene insert. A single CT scan of each patient's limb including the hip, knee and ankle joints was obtained between six weeks and three months postoperatively according to the Perth protocol [12]. Digital images were examined by an independent experienced consultant musculoskeletal radiologist, who was blinded to the intra-operative navigation measurements. The scanning protocol was standardised with the patient supine, the legs in neutral position, knees in full extension and the patellae pointing forwards. Based on the Perth CT protocol [12], coronal alignment of the femoral component was measured. The difference from 90° was recorded as the varus or valgus angle. Femoral component sagittal alignment was measured as the angle between a line from the centre of the femoral head to the centre of the femoral component and a line parallel to the posterior flange of the internal border of the femoral component. Femoral component rotation was measured on the axial images section at the level of epicondyles (identified by the radiologist) as the angle between a line passing through the epicondyles and a line parallel to the posterior aspect of the femoral condyles of the component (Fig. 2 — right). For coronal alignment of the tibial component, the difference from 90° was recorded as the varus or valgus angle. Tibial component sagittal alignment (slope) was measured as the angle between a line passing from the centre of the ankle mortise to centre of tibial component and a line perpendicular to the tibial base plate component. Coronal limb alignment was measured as the angle between a line from centre of femoral head to the centre of the femoral component and a line from the centre of the tibial component to the centre of the ankle
Fig. 1. Example of tibial (top) and femoral (bottom) bone cut validation produced by the imageless navigation system during total knee arthroplasty.
Please cite this article as: Dahabreh Z, et al, Lack of agreement between computer navigation and post-operative 2-dimensional computed tomography (CT) measurements for component..., Knee (2015), http://dx.doi.org/10.1016/j.knee.2015.07.009
Z. Dahabreh et al. / The Knee xxx (2015) xxx–xxx
3
Fig. 2. Example CT image for coronal alignment of the limb and prosthesis components (left) and femoral rotation (right) measurements using the Perth Protocol.
(Fig. 2 — left). All data was entered manually into a spreadsheet during measurement. Varus, component extension, and femoral external rotation were given a positive value. Valgus, component flexion and femoral internal rotation were given a negative value. The difference between CT and navigation measurements for each parameter was calculated for each individual. All continuous data was assessed for normality and equality of variance prior to statistical analysis. Patient age and BMI were compared between SN1 and SN2 using a Mann–Whitney independent samples t-test. The average discrepancy between navigation and CT was assessed whether it was significantly different from zero using Wilcoxon one-sample t-tests. Bland–Altman plots, Typical error [18] and Pearson's correlation were used to assess the agreement between modalities for measurement of femoral and tibial implant positioning. Categorical data such as side of surgery and gender were compared between groups using a Fisher's exact test. The incidence of discrepancies between navigation and CT that exceeded two degree and three degree thresholds (outlier incidence) was calculated for both groups and compared using Fisher's exact test. For alignment measures where no significant differences in incidence were detected, the outlier incidence was calculated for both groups combined and reported as a percentage. Alpha was set a-priori at five percent for all tests and all statistical analyses were performed in Minitab statistical software (version 16, Minitab Inc., MA, USA). 3. Results In a total sample of 50 primary TKAs, nine patients underwent simultaneous bilateral procedures. The 41 patients enrolled were elderly (authors to define if mean ± SD years), predominantly female (23 females) and overweight (BMI 30.1 ± 5.7). There were no significant differences between SN1 and SN2 for age at surgery, side of surgery or BMI. However, SN2 had a significantly (p b 0.05) higher proportion of females to males compared to SN1 (Table 1). Final post-implantation coronal limb alignment measurement with navigation was within three degrees of a neutral (0°) mechanical axis in 96% of cases, compared to 88% when measured by post-operative CT. The differences between navigation and CT measurements for femoral flexion (p b 0.01), femoral rotation (p = 0.013) and tibial slope (p b 0.01) were significantly different to zero (Table 2). Bland–Atlman plots revealed upper and lower limits of agreement ranging from −4.0 to 3.6° (femoral coronal) up to −4.8 to 7.0° (femoral rotation) (Fig. 3). Typical error for component measurements between modalities ranged from 1.4 to 2.1° (Table 2). The incidence of outliers outside two degrees and three degrees of
agreement between navigation and CT measurements ranged between 4.8% for tibial coronal alignment to 82.8% for tibial slope (Table 3). No differences in outlier incidence were observed between SN1 and SN2 for femoral coronal or rotational alignment, coronal tibial alignment or overall coronal limb alignment (Table 3). However, a significant difference (p b 0.01) in outlier incidence was observed between groups for sagittal alignment of the femoral component at the three degrees threshold and for tibial slope at both two degrees and three degrees thresholds (Table 3). Outlier incidence for the two groups combined at the three degrees threshold was between eight percent for tibial coronal alignment and 22% for femoral rotational alignment (Table 4). A similar pattern was observed at the two degrees threshold with a range of 16 to 42% (Table 4). Graphical analysis of the mismatch between navigation and CT for component alignment revealed considerable disagreement for femoral coronal alignment (Fig. 4 — A) and femoral rotation (Fig. 4 — B). Although tibial coronal alignment displayed the lowest incidence of outliers at the two degrees threshold (16%), the range of discrepancies between navigation and CT was also −6 to 6° (Fig. 4 — C). Mismatch was also observed between methods for final coronal alignment, with a higher incidence of outliers at both two degrees and three degrees thresholds than femoral and tibial coronal alignment alone (Fig. 4 — D), although the range was comparable to data for the individual components. Additional analysis of final coronal alignment outliers revealed that 69% of them occurred in TKAs that were within the two degrees threshold for the femoral and the tibial components, 19% in TKAs with outliers in the femoral values, six percent in TKAs with outliers in the tibial component, and six percent in outliers in both femoral and tibial values.
4. Discussion The purpose of this study was to determine the mismatch between intraoperative navigation data and post-operative CT measurements with respect to bone cuts, as well as component and limb alignment following TKA. Combined analysis of SN1 and SN2 data confirmed that Table 1 Patient demographics for S-N 1 and S-N 2.
N (knees) Mean age at surgery Side of surgery BMI (SD) Gender
SN 1
SN 2
p-Value
29 68.6 (50.2–84.1) Right = 17 Left = 12 29.4 (5.5) M = 17 F = 12
21 70.5 (61.1–80) R = 10 L = 11 31.1 (6.1) M=2 F = 19
– 0.38 – 0.35 0.004
Please cite this article as: Dahabreh Z, et al, Lack of agreement between computer navigation and post-operative 2-dimensional computed tomography (CT) measurements for component..., Knee (2015), http://dx.doi.org/10.1016/j.knee.2015.07.009
4
Z. Dahabreh et al. / The Knee xxx (2015) xxx–xxx
Table 2 Validated navigation and post-operative CT measurements (n = 50) of component and limb alignment. Mean difference is calculated by subtracting the navigation measurement from the CT measurement. Navigation (mean ± SD) (°) Femoral Coronal −0.2 (0.6) Flexion −2.7 (1.2) Rotation 1.2 (1.8) Tibial Coronal 0 (0.6) Slope 5.2 (2.0) Limb Alignment Coronal 0.4 (1.6)
CT (mean ± SD) (°)
Mean (±SD) difference (°)
One-sample t-test p-value
Typical error (95% CI) (°)
Pearson correlation
0 (1.8) −1.0 (2.0) 0.1 (1.8)
0.2 (1.9) 1.7 (2.1) −1.1 (3.0)
0.46 b0.01 0.013
1.4 (1.1–1.7) 1.5 (1.2–1.9) 2.1 (1.8–2.7)
0.01 0.20 −0.44
0.1 (1.9) 2.5 (2.5)
0.1 (1.9) −2.7 (2.9)
0.74 b0.01
1.4 (1.1–1.7) 2.1 (1.7–2.6)
0.05 0.17
0.7 (1.8)
0.3 (2.6)
satisfactory outlier incidence (± 3°) from neutral coronal limb alignment was achieved when assessed by both computer navigation (96%) and postoperative CT (88%); results that are comparable to previous reports [6,19]. However, intraoperative measurements did not match post-operative CT values in 8–42% of cases, depending on the component alignment measure and the threshold of acceptable deviation. Mismatch in overall limb alignment could not be attributed to a single component mismatch factor. The reported mismatch between modalities (computer navigation and 2D-CT) on a patient-by-patient basis has important implications for interpretation of alignment using 2D-CT following navigated TKA.
0.361
1.8 (1.6–2.2)
−0.1
Although CT imaging as a modality is accurate and detailed, points of reference that are utilised for further measurement and analysis of alignment are determined by experienced assessors, thus allowing for a certain level subjectivity. Further, the intra and inter-observer reliability of both modalities requires further study and should ideally be reported in subsequent work in this area. Indeed, the question remains regarding a gold standard to determine alignment, with promising results for 3-dimensional CT (3D-CT), showing significantly higher interobserver correlation than 2D-CT for femoral component rotation (0.91 vs 0.29), femoral coronal alignment (0.97 vs 0.65), and tibial component coronal alignment (0.89 vs 0.70) [8].
Fig. 3. Bland–Altman plots of agreement between navigation and CT measurements for lower limb coronal alignment, as well as component alignment in sagittal, coronal and rotation planes. Upper and lower limits are indicated with dashed lines above and below the mean.
Please cite this article as: Dahabreh Z, et al, Lack of agreement between computer navigation and post-operative 2-dimensional computed tomography (CT) measurements for component..., Knee (2015), http://dx.doi.org/10.1016/j.knee.2015.07.009
Z. Dahabreh et al. / The Knee xxx (2015) xxx–xxx Table 3 Outlier incidence at two and three degree thresholds for SN1 (n = 29) and SN2 (n = 21). 2° threshold S-N 1
Femoral Coronal (%) Flexion (%) Rotation (%) Tibial Coronal (%) Slope (%) Limb alignment Coronal (%)
3° threshold
S-N 2
S-N 1 p-Value (Fisher's exact test)
8 (27.6) 2 (9.5) 0.16 10 (34.5) 10 (47.6) 0.39 13 (44.8) 8 (38.1) 0.77
S-N 2
p-Value (Fisher's exact test)
4 (13.8) 2 (9.5) 1.0 3 (10.3) 8 (38.1) 0.04 6 (20.7) 5 (23.8) 1.0
7 (24.1) 24 (82.8)
1 (4.8) 0.12 6 (28.6) 0.0002
11 (37.9)
5 (23.8) 0.37
4 (13.8) 0 (0) 0.13 20 (69.0) 4 (19.1) 0.0006 7 (24.1) 2 (9.5)
0.27
Table 4 Mismatch between CT and computer navigation measurements (n = 50).
Femoral coronal Femoral rotation Tibia coronal Coronal limb alignment
Within 2° (%)
Out of 2° (%)
Within 3° (%)
Out of 3° (%)
40 (80) 29 (58) 42 (84) 34 (68)
10 (20) 21 (42) 8 (16) 16 (32)
44 (88) 39 (78) 46 (92) 41 (82)
6 (12) 11 (22) 4 (8) 9 (18)
Causes for discrepancies between navigation and CT measurements are multifactorial. Errors in navigation (e.g. landmark registration) and CT measurements co-vary and in some cases may be additive with respect to mismatch. Intra-operative femoral and tibial component alignments reflect the validated cut surface of the bone prior to cementation and implantation of prosthetic components. Therefore, mismatch between measurements for component alignment in the coronal and sagittal planes may be partly explained by changes in component alignment
5
following cementation and implantation, especially in the sagittal plane [20]. Points of reference in the femur and the tibia that define axes are also not standardised between navigation and CT. Navigation referencing is based on anatomical landmarks, whereas CT may be based on measured points of the tibial plateau diameters and visual landmarks in the distal femur (highest point of the notch) with the implants in-situ [12]. These factors will certainly have unequal net effects on discrepancies. Other factors may contribute to specific component measurement mismatch. For instance, CT measurement of femoral component sagittal alignment relies on referencing off the posterior condyle of the implant, thus potentially less accurate than coronal measurements of the component. Metal scatter on CT images may also reduce accuracy of measurements particularly in relation to the femoral component. Generally, several 2D-CT based points of reference are more prone to subjective bias than other well defined parameters. Sagittal distal femoral ‘highest point of the notch’ or epicondylar points to determine the transepicondylar axis may be less accurate than identifying the centre of the femoral head using digitised circles. The rotational position of the limb in the CT scanner could also influence the measured angles for component alignment and final limb alignment [21,22]. Tibial slope mismatch was high for both SN-1 and SN-2, with the rate of mismatch higher for SN-1. This may be attributed to the different desired intra-operative slope of the tibial cut which is higher in the SN-1 group, making post-operative CT measurement potentially less accurate if the limb is rotated inside the scanner. The difference in slope of the tibial component may also translate into mismatch in the coronal plane due to parallax error if the limb is rotated and/or flexed [21]. The high incidence of mismatch in femoral component rotation provides insight into potential causative factors, including intra-operative inaccuracy during registration of the epicondyles, or measurement error secondary to subjectively determining the epicondyles on CT. Firstly, the accuracy of determining the femoral rotation axis intraoperatively remains problematic, despite the use of navigation. That is, the surTEA is generated from the lateral and medial epicondyles that are manually identified by the surgeon, with undefined accuracy
Fig. 4. Mismatch between computer navigation and post-operative CT measurements for femoral coronal alignment (A), femoral rotation (B), tibial coronal alignment (C), and for final coronal alignment (D).
Please cite this article as: Dahabreh Z, et al, Lack of agreement between computer navigation and post-operative 2-dimensional computed tomography (CT) measurements for component..., Knee (2015), http://dx.doi.org/10.1016/j.knee.2015.07.009
6
Z. Dahabreh et al. / The Knee xxx (2015) xxx–xxx
[23–25]. Secondly, CT interpretation of femoral rotation is based on points and axes that are subject to inter- and intra-observer variability [26], which has important implications, as it is the basis for all applications of CT data in the clinical setting [27–29]. Variability in 2D-CT may be reduced by utilising 3D-CT imaging [8]. Future research may provide insight into the feasibility of standardising 3D-CT protocols for assessing component and limb alignment in TKA. Coronal tibial component alignment demonstrated the lowest number of outliers outside the two degree (16%) and three degree (eight percent) thresholds. This may be attributed to a combination of more reliably identifiable anatomical landmarks during tibial registration [25] as well as more clearly defined CT measurement points that are less prone to inter- or intra-observer variability [26]. This finding supports the speculation that the higher mismatch observed in other alignment parameters may be related to less reliable reference points. Final limb alignment should be independent of factors causing mismatch of component measurements as it uses reference points which are independent of individual component positioning. Errors in recorded navigation measurements of overall limb alignment may be related to varus or valgus stress that may have been applied passively and unintentionally, when the limb is held extended. Alternatively, rotational positioning of the limb in the CT scanner may lead to a misinterpretation of coronal alignment in a flexed and/or slightly rotated limb [21,22]. Nevertheless, significant discrepancies were found in overall limb alignment that had exceeded those for femoral or tibial component coronal alignment suggesting that mismatch between navigation and CT obtained final limb alignment figures is more likely to be the result of a true lack of correlation between the two modalities. Factors related to rotational limb positioning and of knee flexion may be responsible for the some of the discrepancies. The results of this study should be interpreted in light of some its limitations, which require exploration in future work. Firstly, postoperative CT measurements are based on the implanted prostheses. Small changes in component alignment (coronal and sagittal) might have occurred after preparation of bony surfaces, cementation and then impaction of the definitive prosthesis. A technique which allows validating component alignment after implantation may eliminate these differences and improve comparison. This may be possible with certain tibial component designs but will be much more technically demanding for the femoral component and almost impossible in the femoral sagittal plane. Secondly, the reliability of 2D-CT as well as for computer navigation in TKA is relatively ill-defined, thus comparison of measurements obtained with the two modalities remains challenging. Further, 2D-CT is the most widely available tool used by most surgeons for post-operative assessment of alignment and reflects common practice. However, it may be less accurate for verifying component and limb alignment than 3D-CT especially when evaluating component rotation [8]. Nevertheless, the introduction of 3D-CT is still limited by image scatter from the prosthetic components. To-date, the utility of 3D radiographs has been demonstrated for native segment alignment, but not for prosthetic components [30,31]. Lastly, the use of two separate combinations of surgeons, navigation systems and TKA implants may appear as a limitation, given the differences in tibial slope between implants. However, final intraoperative bone cut verification and alignment measurements should be independent of surgical technique, as would be subsequent CT analysis. 5. Conclusion This study shows that significant mismatch exists between intraoperative navigation and postoperative CT obtained measurements for component and limb alignment following TKA. Mismatch was highest for femoral rotation and lowest for tibial component coronal alignment. A surgeon should consider a multitude of factors when analysing component and limb alignment postoperatively. This is especially true when surgical decision making is based on CT results for rotational alignment of the femoral component or even for overall coronal limb
alignment. A method to perfectly determine component alignment for TKA remains elusive. Additional factors also complicate comparison and replication of this work, such as the surgeon, the geometry of arthroplasty components, and the specific navigation system. Future work should consider controlling as many of these factors before comparing between studies. Conflict of interest Ziad Dahabreh FRCS (Tr & Orth). No conflict of interest. Corey J. Scholes PhD. No conflict of interest. Bruno Giuffre DRANZCR. No conflict of interest. Myles Coolican FRACS. Institutional and Research support from Zimmer. David A. Parker FRACS. Institutional and Research support from Stryker and Smith and Nephew. Acknowledgements The authors wish to acknowledge the assistance of Amy Brierley and Joe Lynch at the Sydney Orthopaedic Research Institute and the staff of North Shore Radiology and Nuclear Medicine with the patient recruitment and data collection. This study was supported financially by the Sydney Orthopaedic Research Institute in-kind support, by Zimmer Inc. and Smith & Nephew. References [1] Kim Y-H, Park J-W, Kim J-S, Park S-D. The relationship between the survival of total knee arthroplasty and postoperative coronal, sagittal and rotational alignment of knee prosthesis. Int Orthop 2014;38:379–85. [2] Parratte S, Pagnano MW, Trousdale RT, Berry DJ. Effect of postoperative mechanical axis alignment on the fifteen-year survival of modern, cemented total knee replacements. J Bone Joint Surg Am 2010;92:2143–9. [3] Gallo J, Goodman SB, Konttinen YT, Wimmer MA, Holinka M. Osteolysis around total knee arthroplasty: a review of pathogenetic mechanisms. Acta Biomater 2013;9: 8046–58. [4] Berger RA, Crossett LS, Jacobs JJ, Rubash HE. Malrotation causing patellofemoral complications after total knee arthroplasty. Clin Orthop Relat Res 1998:144–53. [5] Lustig S, Scholes CJ, Stegeman TJ, Oussedik S, Coolican MR, Parker DA. Sagittal placement of the femoral component in total knee arthroplasty predicts knee flexion contracture at one-year follow-up. Int Orthop 2012;36:1835–9. [6] Hetaimish BM, Khan MM, Simunovic N, Al-Harbi HH, Bhandari M, Zalzal PK. Metaanalysis of navigation vs conventional total knee arthroplasty. J Arthroplasty 2012; 27:1177–82. [7] Zamora LA, Humphreys KJ, Watt AM, Forel D, Cameron AL. Systematic review of computer-navigated total knee arthroplasty. ANZ J Surg 2013;83:22–30. [8] Hirschmann MT, Konala P, Amsler F, Iranpour F, Friederich NF, Cobb JP. The position and orientation of total knee replacement components: a comparison of conventional radiographs, transverse 2D-CT slices and 3D-CT reconstruction. J Bone Joint Surg 2011;93:629–33. [9] Kannan A, Hawdon G, McMahon SJ. Effect of flexion and rotation on measures of coronal alignment after TKA. J Knee Surg 2012;25:407–10. [10] Skytta ET, Lohman M, Tallroth K, Remes V. Comparison of standard anteroposterior knee and hip-to-ankle radiographs in determining the lower limb and implant alignment after total knee arthroplasty. Scand J Surg 2009;98:250–3. [11] Blakeney WG, Khan RJ, Wall SJ. Computer-assisted techniques versus conventional guides for component alignment in total knee arthroplasty: a randomized controlled trial. J Bone Joint Surg Am 2011;93:1377–84. [12] Chauhan SK, Clark GW, Lloyd S, Scott RG, Breidahl W, Sikorski JM. Computer-assisted total knee replacement. A controlled cadaver study using a multi-parameter quantitative CT assessment of alignment (the Perth CT Protocol). J Bone Joint Surg 2004; 86:818–23. [13] Konigsberg B, Hess R, Hartman C, Smith L, Garvin K. Inter- and Intraobserver Reliability of Two-dimensional CT Scan for Total Knee Arthroplasty Component Malrotation. Clin Orthop Relat Res® 2014;472(1):212–7. [14] Lustig S, Fleury C, Goy D, Neyret P, Donell ST. The accuracy of acquisition of an imageless computer-assisted system and its implication for knee arthroplasty. Knee 2011;18:15–20. [15] Zhang GQ, Chen JY, Chai W, Liu M, Wang Y. Comparison between computerassisted-navigation and conventional total knee arthroplasties in patients
Please cite this article as: Dahabreh Z, et al, Lack of agreement between computer navigation and post-operative 2-dimensional computed tomography (CT) measurements for component..., Knee (2015), http://dx.doi.org/10.1016/j.knee.2015.07.009
Z. Dahabreh et al. / The Knee xxx (2015) xxx–xxx
[16]
[17]
[18] [19]
[20]
[21] [22] [23]
undergoing simultaneous bilateral procedures: a randomized clinical trial. J Bone Joint Surg Am 2011;93:1190–6. Doro LC, Hughes RE, Miller JD, Schultz KF, Hallstrom B, Urquhart AG. The reproducibility of a kinematically-derived axis of the knee versus digitized anatomical landmarks using a knee navigation system. Open Biomed Eng J 2008;2:52–6. Widmer BJ, Scholes CJ, Lustig S, Conrad L, Oussedik SI, Parker DA. Intraoperative computer navigation parameters are poor predictors of function 1 year after total knee arthroplasty. J Arthroplasty 2013;28:56–61. Hopkins WG. Measures of reliability in sports medicine and science. Sports Med 2000;30:1–15. Cheng T, Zhao S, Peng X, Zhang X. Does computer-assisted surgery improve postoperative leg alignment and implant positioning following total knee arthroplasty? A meta-analysis of randomized controlled trials? Knee Surg Sports Traumatol Arthrosc 2012;20:1307–22. Catani F, Biasca N, Ensini A, Leardini A, Bianchi L, Digennaro V, et al. Alignment deviation between bone resection and final implant positioning in computernavigated total knee arthroplasty. J Bone Joint Surg Am 2008;90:765–71. Lonner JH, Laird MT, Stuchin SA. Effect of rotation and knee flexion on radiographic alignment in total knee arthroplasties. Clin Orthop Relat Res 1996:102–6. Radtke K, Becher C, Noll Y, Ostermeier S. Effect of limb rotation on radiographic alignment in total knee arthroplasties. Arch Orthop Trauma Surg 2010;130:451–7. Robinson M, Eckhoff DG, Reinig KD, Bagur MM, Bach JM. Variability of landmark identification in total knee arthroplasty. Clin Orthop Relat Res 2006;442:57–62.
7
[24] Yau WP, Chiu KY, Tang WM. How precise is the determination of rotational alignment of the femoral prosthesis in total knee arthroplasty: an in vivo study. J Arthroplasty 2007;22:1042–8. [25] Yau WP, Leung A, Liu KG, Yan CH, Wong LL, Chiu KY. Interobserver and intraobserver errors in obtaining visually selected anatomical landmarks during registration process in non-image-based navigation-assisted total knee arthroplasty. J Arthroplasty 2007;22:1150–61. [26] Victor J, Van Doninck D, Labey L, Innocenti B, Parizel PM, Bellemans J. How precise can bony landmarks be determined on a CT scan of the knee? Knee 2009;16:358–65. [27] Amiot LP, Poulin F. Computed tomography-based navigation for hip, knee, and spine surgery. Clin Orthop Relat Res 2004:77–86. [28] Jazrawi LM, Birdzell L, Kummer FJ, Di Cesare PE. The accuracy of computed tomography for determining femoral and tibial total knee arthroplasty component rotation. J Arthroplasty 2000;15:761–6. [29] Nishihara S, Sugano N, Ikai M, Sasama T, Tamura Y, Tamura S, et al. Accuracy evaluation of a shape-based registration method for a computer navigation system for total knee arthroplasty. J Knee Surg 2003;16:98–105. [30] Folinais D, Thelen P, Delin C, Radier C, Catonne Y, Lazennec JY. Measuring femoral and rotational alignment: EOS system versus computed tomography. Orthop Traumatol Surg Res 2013;99:509–16. [31] Guenoun B, Zadegan F, Aim F, Hannouche D, Nizard R. Reliability of a new method for lower-extremity measurements based on stereoradiographic three-dimensional reconstruction. Orthop Traumatol Surg Res 2012;98:506–13.
Please cite this article as: Dahabreh Z, et al, Lack of agreement between computer navigation and post-operative 2-dimensional computed tomography (CT) measurements for component..., Knee (2015), http://dx.doi.org/10.1016/j.knee.2015.07.009