The Journal of Arthroplasty 33 (2018) 1572e1578

Contents lists available at ScienceDirect

The Journal of Arthroplasty

journal homepage: www.arthroplastyjournal .org

Basic Science

Interobserver and Intraobserver Reliability of ComputedTomographyeBased Three-Dimensional Preoperative Planning forPrimary Total Knee Arthroplasty

Michiaki Miura, MD *, Shigeo Hagiwara, MD, Junichi Nakamura, MD, Yasushi Wako, MD,Yuya Kawarai, MD, Seiji Ohtori, MDDepartment of Orthopedic Surgery, Graduate School of Medicine, Chiba University, Chiba, Japan

a r t i c l e i n f o

Article history:Received 8 June 2017Received in revised form15 December 2017Accepted 18 December 2017Available online 28 December 2017

Keywords:total knee arthroplastypreoperative planningCT-based planningreliabilityZed knee

Junichi Nakamura was supported by JSPS KAKENHIthe other authors declared no conflicts of interest in

One or more of the authors of this paper have discconflicts of interest, which may include receipt of payminstitutional support, or association with an entity inmay be perceived to have potential conflict of intedisclosure statements refer to https://doi.org/10.1016/* Reprint requests: Michiaki Miura, MD, Departm

Graduate School of Medicine, Chiba University, 1-8Chiba Prefecture 260-8677, Japan.

https://doi.org/10.1016/j.arth.2017.12.0350883-5403/© 2017 Elsevier Inc. All rights reserved.

a b s t r a c t

Background: Preoperative planning is an important factor for total knee arthroplasty (TKA). The aim ofthis study is to document the interobserver and intraobserver reliability of computed tomography (CT)-based 3-dimensional (3D) preoperative planning for primary TKA.Methods: Twenty knees (10 with osteoarthritis and 10 with rheumatoid arthritis) were studied inde-pendently by 6 orthopedic surgeons using a CT-based 3D planning system. The measurements weremade twice at more than 3-week intervals without any knowledge of their own previous measurementsor those of the others. We assessed the femoral and tibial component sizes and the alignment of thefemoral component.Results: The interobserver and intraobserver agreements for femoral component size were 44.3% and62.5% with exact size, and increased to 90.7% and 99.2% within one size difference; the intraclasscorrelation coefficients (ICCs) were 0.919 and 0.936, respectively. The interobserver and intraobserveragreements for tibial component size were 57.0% and 66.7% with exact size, and increased to 87.3% and90.0% within one size difference; the ICCs were 0.909 and 0.924, respectively. The ICCs for femoral andtibial size were better in rheumatoid arthritis than in osteoarthritis. Interobserver ICC for femoral valgusangle was 0.807, and 0.893 for intraobserver reliability. Interobserver ICC of the femoral external rotationangle was 0.463, and 0.622 for intraobserver reliability.Conclusion: CT-based 3D preoperative planning for primary TKA has clinical implications for predictingappropriate size and alignment of the component in patients with osteoarthritis and rheumatoidarthritis.

© 2017 Elsevier Inc. All rights reserved.

Inappropriate size, malposition, and malalignment of thecomponents have led to poor outcomes of total knee arthroplasty(TKA) [1e3]. Therefore, many researchers have tried to improve the

grant number 17K10954 andthis study.

losed potential or pertinentent, either direct or indirect,the biomedical field which

rest with this work. For fullj.arth.2017.12.035.ent of Orthopedic Surgery,-1 Inohana, Chuo-ku, Chiba,

reliability of their preoperative planningmethods [4e7]. Traditional2-dimensional (2D) planning has been the gold standard interna-tionally by placing translucent templates on standard anterior-posterior and lateral radiographs of the knee [8,9]. Several studieshave revealed the advantage of digital 2D planning for predictingimplant size [6,7], but in some cases there are limitations to thecorrect sizing of the component using only radiographs because ofskeletal deformities. Recently, computed tomography (CT)ebased3-dimensional (3D) planning has been shown to more accuratelyestimate the size as well as the alignment and rotation of thecomponents [7,10e14]. Previous studies compared preoperativeplanningwith postoperative actual size and alignment, but they didnot evaluate the precision of the planning method itself [4e7].Validation of the preoperative planning method is essential beforecomparing postoperative outcomes. To our knowledge, only one

Table 1Subject Characteristics.

OA RA P Value

Number of joints 10 knees 10 kneesAge (y) (range) 76.9 (66-83) 60.7 (39-76) .006a

Male:female 2:8 2:8 1b

BMI 27.4 (23.3-31.6) 23.4 (17.6-27.6) .035a

FTA (range) 182 (176-187) 174 (162-185) .030a

BMI, body mass index; FTA, femorotibial angle.a Mann-Whitney U-test.b Chi-squared test.

M. Miura et al. / The Journal of Arthroplasty 33 (2018) 1572e1578 1573

article showed validation of CT-based 3D preoperative planning fortotal hip arthroplasty [15]. The aim of this study is to document theinterobserver and intraobserver reliabilities of CT-based 3Dpreoperative planning for primary TKA.

Materials and Methods

The research protocol of this prospective study was approved bythe institutional review boards and all participants gave theirwritten informed consent. We followed the Guidelines forReporting Reliability and Agreement Studies [16].

We analyzed 20 knees from 14 patients (2 men and 12women) with no history of previous knee surgery who werescheduled for primary TKA in our institution. A sample size of 20was chosen in order to have sufficient variability of radiologicalfindings without inducing fatigue and loss of attention amongthe observers, according to the statistical methods of Herringet al [17] and Nakamura et al [18]. The mean age of the patientswas 68.8 years (range 39-83, standard deviation 13.0). Ten knees

Fig. 1. 3D simulation of TKA on the right knee. The Zed Knee system created the 3D bone mgave visual information of reposition to the surgeons in advance. Frontal view (A) and late

with osteoarthritis (OA) all classified as Kellgren-Lawrence grade4, and 10 knees with rheumatoid arthritis (RA), were included(Table 1).

All patients had CT scans taken from the hip joint to the anklejoint using Aquilion ONE (TOSHIBA Medical Systems Co, Tokyo,Japan), with a slice thickness of 1 mm. The data were importedinto the Zed Knee System (LEXI, Tokyo, Japan). The Zed KneeSystem is a CT-based 3D preoperative planning softwareprogram for TKA. We used a cruciate-retaining implant (FineKnee CR; Teijin Nakashima Medical, Okayama, Japan) and plan-ned to fix the components with cement for all patients. Thecomponents were available in 6 sizes for both femoral and tibialsides. First, each observer sets several bony landmarks tocalibrate the individual anatomy and made a 3D reconstructionmodel (Fig. 1). Bony landmarks of the femur were the center ofthe femoral head and several points on the distal femur (Fig. 2).Tibial landmarks were the shaft axis, the anterior-posterior axis,and the articular surface (Fig. 3). Second, we chose the size,position, and alignment of the component according to surgicalparameters and target (Table 2). The femoral component was setto avoid overhang or notching, and the rotation of the compo-nent was determined to follow the surgical epicondylar axis inthe axial plane. The coronal alignment of the component was setvertical to the line connecting the hip center and the center ofthe intercondylar fossa, and sagittal alignment was setvertical to the femoral shaft axis. The tibial component was alsoset to exclude overhang or notching, and its rotation was fixed tofollow the anterior-posterior axis of the “Akagi line” (the axisfrom the medial third of the tibial tubercle to the enthesis of theposterior cruciate ligament) [19]. Its sagittal and coronal align-ments were fixed vertically to the tibial shaft axis. The femoralvalgus angle (FVA) was defined as the angle between the distal

odel with femoral and tibial components in accordance with the planning rule, whichral view (B).

Fig. 2. Bony landmarks of the distal end of the right femur. (A) Coronal slice of the knee center shows the most distal points of the medial (arrow) and lateral (arrow head) condyle.(B) Axial slice of the femoral epicondyle shows the most posterior points of the medial (*) and lateral (y) condyle, and the medial (¶) and lateral (x) epicondyle. (C) Sagittal slice of themedical femoral condyle. (D) Sagittal slice of the lateral femoral condyle.

M. Miura et al. / The Journal of Arthroplasty 33 (2018) 1572e15781574

femoral shaft axis and the line vertical to the component surfacein the coronal plane (Fig. 4). The femoral external rotation angle(FERA) was defined as the angle between the axis of the femoralcomponent and the posterior condylar axis in the axial plane(Fig. 4).

Statistical Analysis

Preoperative planning was independently performed by 6 boardcertified orthopedic surgeons with 8-16 years of experience. Inorder to assess the intraobserver reliability, the trials were donetwice at more than 3-week intervals without any knowledge of theobserver’s own previous measurements or those of the others. Thepercentage agreement with the exact femoral and tibial componentsize and within one size difference was assessed, as was theintraclass correlation coefficient (ICC). We also assessed the gapbetween the 2 trials and ICC for FVA and FERA for each subject.Interobserver and intraobserver reliabilities were determined withICC (2, 1) and ICC (1, 1), respectively. The scoring system of Landiswas used for the analysis (almost perfect: >0.81, substantial:0.61-0.80, moderate: 0.41-0.60, fair: 0.21-0.40, slight: 0.0-0.20)[20]. The Mann-Whitney U-test was used to compare reliabilitybetween OA and RA patients. The data were analyzed usingBellCurve for Excel (Social Survey Research Information Co., Ltd.)

for Windows and 95% confidence intervals for ICC values areprovided. A P-value <.05 was considered statistically significant.

Results

Interobserver Reliability

Femoral Component SizeThe mean percentage agreement of the femoral component size

was 44.3% with the exact size and increased to 90.7% within onesize difference (Table 3). The ICC of femoral component size showedalmost perfect agreement, and the ICC for RA patients was higherthan that for OA patients (Table 4).

Tibial Component SizeThe mean percentage agreement of tibial component size was

57.0% with the exact size and increased to 87.3% within one sizedifference (Table 3). The ICC of tibial component size showedalmost perfect agreement, and the ICC for RA patients was higherthan that for OA patients (Table 4).

Femoral Valgus AngleThe mean difference in the FVA between each pair of observers

was 1.0�, ranging from 0.5� to 1.7�. The ICC of the FVA showed

Fig. 3. Bony landmarks of the right tibia. (A) Coronal slice of the tibia shows the medial (arrow) and lateral (arrow head) articular surfaces. Tibial shaft axis is determined by themedullary centers at proximal one-third (*) and at distal one-third (y). (B) Sagittal slice of the tibia shows the medial one-third of the tibial tuberosity (¶) and the enthesis of theposterior cruciate ligament (x), constituting the anteroposterior axis “Akagi line”. (C) Sagittal slice of the medical tibial condyle. (D) Sagittal slice of the lateral tibial condyle.

Table 2Surgical Parameters and Target.

Parameter Assess/Fixed

Target

Component sizeFemur Assess Bone coverage without overhang or notchingTibia Assess Bone coverage without overhang or notching

Coronal alignmentFemur Assess Vertical to the line connecting the hip

center and the center of the condylesTibia Fixed 90� to the tibial shaft axis

Sagittal alignmentFemur Fixed 90� to the femoral shaft axisTibia Fixed 90� to the tibial shaft axis

Rotational alignmentFemur Assess Following the surgical epicondylar axisTibia Fixed Following the Akagi line

M. Miura et al. / The Journal of Arthroplasty 33 (2018) 1572e1578 1575

substantial agreement, and there was no significant differencebetween OA and RA patients (Table 4).

Femoral External Rotation AngleThe mean difference in the FERA between each pair of the

observers was 1.8�, ranging from 1.0� to 2.4�. The ICC of the FERAshowed moderate agreement, and there was no significant differ-ence between OA and RA patients (Table 4).

Intraobserver Reliability

Femoral Component SizeThe mean percentage agreement of femoral component size in

the 2 trials of each observer was 62.5% with the exact size. Thevalue increased to 99.2% within one size difference (Table 5). TheICC of femoral component size showed almost perfect agreement,and the ICC for RA patients was higher than that for OA patients(Table 6).

Tibial Component SizeThe mean percentage agreement of tibial component size in

the 2 trials of each observer was 66.7% with the exact size. Thevalue increased to 90.0% within one size difference (Table 5). TheICC of tibial component size showed almost perfect agreement,and there was no significant difference between OA and RApatients (Table 6).

Femoral Valgus AngleThe mean intraobserver difference in the FVA was 0.7�, ranging

from 0.5� to 0.9�. The ICC of the FVA showed almost perfectagreement, and therewas no significant difference between OA andRA patients (Table 6).

Femoral External Rotation AngleThe mean intraobserver difference in the FERAwas 1.4�, ranging

from 0.6� to 2.4�. The ICC of the FERA showed substantial

Fig. 4. The definition of FVA and FERA. From a coronal slice through the center of the femoral head (A), the FVA is the angle between the distal femoral shaft axis (a) and the linevertical to the component surface (b). From an axial slice at the femoral epicondyle (B), the FERA is the angle between the axis of the femoral component and the posterior condylaraxis (c). The surgical epicondylar axis is referenced (d).

M. Miura et al. / The Journal of Arthroplasty 33 (2018) 1572e15781576

agreement, and therewas no significant difference between OA andRA patients (Table 6).

Table 3Interobserver Percent Agreement of Component Size.

Overall (n ¼ 20) FemoralComponent Size

TibialComponent Size

Exact sizeMean (SD) 44.3% (18.6) 57.0% (17.4)CI 34.9-53.7 48.2-65.8

One size differenceMean (SD) 90.7% (11.6) 87.3% (17.0)CI 84.8-96.6 78.7-95.9

Diagnosis (n ¼ 10) OA RA OA RA

Exact sizeMean (SD) 42.0% (24.3) 46.7% (21.3) 53.3% (21.9) 60.7% (17.1)CI 29.7-54.3 35.9-57.5 42.2-64.4 52.0-69.4U valuea 106.5 89.0P valuea .801 .324

One size differenceMean (SD) 89.3% (15.8) 92.0% (10.1) 84.0% (23.7) 90.7% (11.0)CI 81.3-97.3 86.9-97.1 72.0-96.0 85.1-96.3U valuea 111 111.0P valuea .946 .947

SD, standard deviation; CI, confidence intervals.a Mann-Whitney U-test.

Discussion

This is the first study to report the interobserver and intra-observer reliability of 3D preoperative planning software forTKA (Zed Knee System; LEXI). Interobserver and intraobserveragreement of the component size was about 50% for exact sizeand increased to about 90% within one size difference. McLaw-horn et al [4] reviewed the literature comparing component sizebased on preoperative planning with actual component sizepostoperatively, and reported that accuracy was 63.2% for exactsize and 97.9% within one size difference of the femur. For thetibia, accuracy was 62.6% for exact size of the component and96.8% within one size difference. Thus, our results wereconsistent with McLawhorn’s review, although it should benoted that the methods in each study were different. The 2studies varied for number of observers, sample size, kind ofimplant, and planning methods by radiograph, axial 2D or 3Dreconstructed CT [4]. The authors concluded that planning thetarget within one size difference was acceptable because theintraoperative final decision should be at the surgeon’sdiscretion [4,5,7,13].

Some previous studies have described interobserver andintraobserver reliability for preoperative 2D planning ofcomponent sizes for TKA. The et al reported the interobserverreliability (k value ¼ 0.63-0.75) and intraobserverreliability (k value ¼ 0.82-0.88) for 2D planning [8]. Hsu et al [9]also reported the interobserver (ICC ¼ 0.86) and intraobserverreliability (k value ¼ 0.90) with 2D planning. We achieved higher

interobserver (ICC ¼ 0.909-0.919) and intraobserver reliability(ICC ¼ 0.924-0.936) for component sizes with CT-based 3Dplanning. The reliability of component sizes for RA patientstended to be better than those for OA patients. These differencescould be attributed to the fact that all the OA patients wereclassified into Kellgren-Lawrence grade 4, and the bone defor-mity and large osteophytes were more severe in some OApatients than in RA patients.

Table 4Interobserver Reliability of Intraclass Correlation Coefficient.

Overall (n ¼ 20) F Size T Size FVA FERA

Mean (SD) 0.919 (0.040) 0.909 (0.064) 0.807 (0.163) 0.463 (0.191)CI 0.797-0.968 0.769-0.964 0.555-0.923 0.274-0.781

Diagnosis (n ¼ 10) OA RA OA RA OA RA OA RA

Mean (SD) 0.867 (0.081) 0.938 (0.035) 0.878 (0.066) 0.929 (0.080) 0.774 (0.083) 0.626 (0.291) 0.378 (0.236) 0.449 (0.306)CI 0.416-0.934 0.608-0.976 0.399-0.867 0.678-0.979 0.359-0.944 0.289-0.947 0.102-0.799 0.112-0.880U valuea 47.000 55.000 97.00 96.00P valuea .007 .017 .520 .494

F size, femoral component size; T size, tibial component size; SD, standard deviation; CI, confidence intervals.a Mann-Whitney U-test.

Table 5Intraobserver Percent Agreement of Component Size.

Overall (n ¼ 20) Femoral Component Size Tibial Component Size

Exact sizeMean (SD) 62.5% (21.6) 66.7% (10.8)CI 45.2-79.8 58.1-75.3

One size differenceMean (SD) 99.2% (2.0) 90.0% (13.8)CI 97.6-100 79.0-100

Diagnosis (n ¼ 10) OA RA OA RA

Exact sizeMean (SD) 58.3% (28.6) 66.7% (21.6) 68.3% (21.4) 65.0% (10.5)CI 35.4-81.2 49.4-84.0 51.2-85.4 56.6-73.4U valuea 14.5 13.0P valuea .572 .413

One size differenceMean (SD) 98.3% (4.1) 100% (0) 90.0% (15.5) 90.0% (12.6)CI 95.0-100 100-100 77.6-100 79.9-100U valuea 15.0 17.0P valuea .317 .857

SD, standard deviation; CI, confidence intervals.a Mann-Whitney U-test.

M. Miura et al. / The Journal of Arthroplasty 33 (2018) 1572e1578 1577

The advantage of 3D planning over 2D planning is the rota-tional alignment. The current literature suggests that the bestmethod for assessing component rotation is a CT scan [6,21].Nevertheless, there have been few studies of the reliability ofTKA component rotation using CT scans. Konigsberg et al [14]described the interobserver (ICC ¼ 0.386) and intraobserverreliability (ICC ¼ 0.606) of femoral component rotation in pa-tients with TKA based on measurements by 2 orthopedic sur-geons and 1 radiologist. Hirschmann et al [6] reportedinterobserver ICC of 0.29 and intraobserver ICC of 0.60 for FERAwith 2D CT, and improved interobserver ICC of 0.91 and intra-observer ICC of 0.73 for FERA with 3D CT. Park et al [12] alsoreported interobserver ICC of 0.66-0.67 and intraobserver ICC of0.76-0.97 with magnetic resonance imaging. Jazrawi et al [10]showed an excellent correlation of 0.85-0.93 with 2D CT in acadaveric study. Our results (interobserver ICC of 0.463 andintraobserver ICC of 0.622) were similar to Konigsberg’s studybut lower than those of other studies. Park et al [12] stated thatanatomic landmarks such as the femoral epicondyles to definefemoral component rotation have high interobserver variability.Kobayashi et al [22] also claimed that the severity of the osteo-phytes and deformity of the distal femur made it difficult toidentify the medial sulcus and the posterior condylar axis. In fact,OA patients with severe deformity showed poor reliability forestimates of component rotation.

A major limitation of our study was the lack of comparisonbetween the preoperatively planned and actual components.Thus, the described preoperative templating method was “pre-cise” (multiple observers report similar findings at repeated timepoints), but this was independent of “accuracy,” which wouldrequire a comparison to intraoperative or postoperative findings.However, several authors have already reported such compara-tive studies in 3D planning for TKA [4e7,10,11], so we focused onthe interobserver and intraobserver reliability of CT-based 3Dpreoperative planning. Second, it is controversial whetherpreoperative planning can reduce intraoperative sizing andalignment errors, and can improve outcomes in TKA. Theoutcome of TKA seems to be multifactorial. Smaller tibialcomponent size relative to femoral component size affectsloading patterns in the proximal tibia and leads to loosening ormigration [23,24]. Varus coronal malalignment of the tibialcomponent also causes abnormal load concentrations on thetibia, and can result in early failure of the components [2,24].Component malrotation influences patellofemoral tracking andcan cause early patellar dislocation or late patellar componentfailure [3,25,26]. The chances for a successful outcome for TKA

can be improved by reliable preoperative planning of femoraland tibial component sizes, and coronal and rotational align-ment of the femoral component. The third limitation was thatwe did not evaluate the rotational alignment of the tibia becauseby definition the tibial component was fixed to the “Akagi line”[19] in all cases. The reliability of this line has been reportedwith moderate interobserver ICC (0.53-0.64) and substantialintraobserver ICC (0.67-0.84) [12]. But, the rotation of the tibialcomponent was also determined intraoperatively by severalanatomical landmarks including the anterior tibial crest, thesecond ray, and the first web space of the foot [27,28]. So far,there is no gold standard measurement for tibial componentrotation [29]. Furthermore, a recent study suggests the impor-tance of femorotibial kinematics and the utility of the intra-operative range of motion technique [30].

In conclusion, CT-based 3D preoperative planning for TKAshowed excellent reliability for component size and FVA, butmoderate reliability for FERA in patients with OA and RA. Furtherstudies are needed to show the accuracy of CT-based 3D planning incomparison with postoperative implant size and alignment, andwhether the accurate size and alignment can improve the clinicaloutcome of the TKA.

Table 6Intraobserver Reliability of Intraclass Correlation Coefficient.

Overall (n ¼ 20) F Size T Size FVA FERA

Mean (SD) 0.936 (0.040) 0.924 (0.052) 0.893 (0.130) 0.622 (0.188)CI 0.840-0.974 0.812-0.970 0.735-0.957 0.384-0.849

Diagnosis (n ¼ 10) OA RA OA RA OA RA OA RA

Mean (SD) 0.768 (0.370) 0.960 (0.027) 0.823 (0.220) 0.947 (0.034) 0.856 (0.125) 0.816 (0.093) 0.581 (0.433) 0.557 (0.330)CI 0.736-0.946 0.856-0.991 0.590-0.937 0.789-0.987 0.504-0.966 0.387-0.941 0.402-0.864 0.274-0.865U valuea 5.500 9.000 13.00 14.00P valuea .0453 .1495 .4233 .5218

F size, femoral component size; T size, tibial component size; SD, standard deviation; CI, confidence intervals.a Mann-Whitney U-test.

M. Miura et al. / The Journal of Arthroplasty 33 (2018) 1572e15781578

Acknowledgments

The authors thank Kento Nawata and Masahiko Sugano forparticipating in preoperative planning.

References

[1] Berend ME, Small SR, Ritter MA, Buckley CA, Merk JC, Dierking WK. Effects offemoral component size on proximal tibial strain with anatomic graduatedcomponents total knee arthroplasty. J Arthroplasty 2010;25:58e63.

[2] Ritter MA, Faris PM, Keating EM, Meding JB. Postoperative alignment of totalknee replacement. Its effect on survival. Clin Orthop Relat Res 1994:153e6.

[3] Berger RA, Crossett LS, Jacobs JJ, Rubash HE.Malrotation causing patellofemoralcomplications after total knee arthroplasty. ClinOrthopRelat Res 1998:144e53.

[4] McLawhorn AS, Carroll KM, Blevins JL, DeNegre ST, Mayman DJ, Jerabek SA.Template-directed instrumentation reduces cost and improves efficiency fortotal knee arthroplasty: an economic decision analysis and pilot study.J Arthroplasty 2015;30:1699e704.

[5] Goyal N, Stulberg SD. Evaluating the precision of preoperative planning inpatient specific instrumentation: can a single MRI yield different preoperativeplans? J Arthroplasty 2015;30:1250e3.

[6] Hirschmann MT, Konala P, Amsler F, Iranpour F, Friederich NF, Cobb JP. Theposition and orientation of total knee replacement components: a comparisonof conventional radiographs, transverse 2D-CT slices and 3D-CT reconstruc-tion. J Bone Joint Surg Br 2011;93:629e33.

[7] Kobayashi A, Ishii Y, Takeda M, Noguchi H, Higuchi H, Toyabe S. Comparison ofanalog 2D and digital 3D preoperative templating for predicting implant sizein total knee arthroplasty. Comput Aided Surg 2012;17:96e101.

[8] The B, Diercks RL, van Ooijen PM, van Horn JR. Comparison of analog anddigital preoperative planning in total hip and knee arthroplasties. A pro-spective study of 173 hips and 65 total knees. Acta Orthop 2005;76:78e84.

[9] Hsu AR, Kim JD, Bhatia S, Levine BR. Effect of training level on accuracy ofdigital templating in primary total hip and knee arthroplasty. Orthopedics2012;35:e179e83.

[10] Jazrawi LM, Birdzell L, Kummer FJ, Di Cesare PE. The accuracy of computedtomography for determining femoral and tibial total knee arthroplastycomponent rotation. J Arthroplasty 2000;15:761e6.

[11] Franceschi JP, Sbihi A, Computer Assisted Orthopedic Surgery - France (CAOS -France). 3D templating and patient-specific cutting guides (Knee-Plan) in totalknee arthroplasty: postoperative CT-based assessment of implant positioning.Orthop Traumatol Surg Res 2014;100(6 Suppl):S281e6.

[12] Park A, Nam D, Friedman MV, Duncan ST, Hillen TJ, Barrack RL. Inter-observerprecision and physiologic variability of MRI landmarks used to determinerotational alignment in conventional and patient-specific TKA. J Arthroplasty2015;30:290e5.

[13] Ettinger M, Claassen L, Paes P, Calliess T. 2D versus 3D templating in total kneearthroplasty. Knee 2016;23:149e51.

[14] Konigsberg B, Hess R, Hartman C, Smith L, Garvin KL. Inter- and intraobserverreliability of two-dimensional CT scan for total knee arthroplasty componentmalrotation. Clin Orthop Relat Res 2014;472:212e7.

[15] Wako Y, Nakamura J, Miura M, Kawarai Y, Sugano M, Nawata K. Interobserverand intraobserver reliability of three-dimensional preoperative planningsoftware in total hip arthroplasty. J Arthroplasty 2017. https://doi.org/10.1016/j.arth.2017.08.031. pii: S0883e5403(17)30754-4; [Epub ahead ofprint].

[16] Kottner J, Audige L, Brorson S, Donner A, Gajewski BJ, Hr�objartsson A, et al.Guidelines for Reporting Reliability and Agreement Studies (GRRAS) wereproposed. Int J Nurs Stud 2011;48:661e71.

[17] Herring JA, Kim HT, Browne R. Legg-Calve-Perthes disease. Part I: classifica-tion of radiographs with use of the modified lateral pillar and Stulbergclassifications. J Bone Joint Surg Am 2004;86-A:2103e20.

[18] Nakamura J, Kishida S, Harada Y, Iida S, Oinuma K, Yamamoto S, et al. Inter-observer and intra-observer reliabilities of the Japanese Ministry of Health,Labor and Welfare type classification system for osteonecrosis of the femoralhead. Mod Rheumatol 2011;21:488e94.

[19] Akagi M, Oh M, Nonaka T, Tsujimoto H, Asano T, Hamanishi C. An ante-roposterior axis of the tibia for total knee arthroplasty. Clin Orthop Relat Res2004:213e9.

[20] Landis JR, Koch GG. The measurement of observer agreement for categoricaldata. Biometrics 1977;33:159e74.

[21] Figueroa J, Guarachi JP, Matas J, Arnander M, Orrego M. Is computed tomog-raphy an accurate and reliable method for measuring total knee arthroplastycomponent rotation? Int Orthop 2016;40:709e14.

[22] Kobayashi H, Aratake M, Akamatsu Y, Mitsugi N, Taki N, Saito T. Reproduc-ibility of condylar twist angle measurement using computed tomography andaxial radiography of the distal femur. Orthop Traumatol Surg Res 2014;100:885e90.

[23] Berend ME, Ritter MA, Hyldahl HC, Meding JB, Redelman R. Implantmigration and failure in total knee arthroplasty is related to body massindex and tibial component size. J Arthroplasty 2008;23(6 Suppl 1):104e9.

[24] Berend ME, Ritter MA, Meding JB, Faris PM, Keating EM, Redelman R, et al.Tibial component failure mechanisms in total knee arthroplasty. Clin OrthopRelat Res 2004:26e34.

[25] Barrack RL, Schrader T, Bertot AJ, Wolfe MW, Myers L. Component rotationand anterior knee pain after total knee arthroplasty. Clin Orthop Relat Res2001:46e55.

[26] Matsuda S, Miura H, Nagamine R, Urabe K, Hirata G, Iwamoto Y. Effect offemoral and tibial component position on patellar tracking following totalknee arthroplasty: 10-year follow-up of Miller-Galante I knees. Am J KneeSurg 2001;14:152e6.

[27] Tsukeoka T, Tsuneizumi Y, Lee TH. Accuracy of the second metatarsal asa landmark for the extramedullary tibial cutting guide in totalknee arthroplasty. Knee Surg Sports Traumatol Arthrosc 2014;22:2969e74.

[28] Tsukeoka T, Lee TH, Tsuneizumi Y, Suzuki M. The tibial crest as a practicaluseful landmark in total knee arthroplasty. Knee 2014;21:283e9.

[29] Gromov K, Korchi M, Thomsen MG, Husted H, Troelsen A. What is the optimalalignment of the tibial and femoral components in knee arthroplasty? ActaOrthop 2014;85:480e719.

[30] Ikeuchi M, Yamanaka N, Okanoue Y, Ueta E, Tani T. Determining the rotationalalignment of the tibial component at total knee replacement: a comparison oftwo techniques. J Bone Joint Surg Br 2007;89:45e9.