Secondary Logo

Journal Logo

SECTION I: SYMPOSIUM I: Virtual Reality Symposium

Variability of Landmark Identification in Total Knee Arthroplasty

Robinson, Mitchel*; Eckhoff, Donald, G*; Reinig, Karl, D*; Bagur, Michelle, M*; Bach, Joel, M*†

Section Editor(s): Mabrey, Jay D MD, Guest Editor

Author Information
Clinical Orthopaedics and Related Research: January 2006 - Volume 442 - Issue - p 57-62
doi: 10.1097/01.blo.0000197081.72341.4b


Virtual images have numerous applications in orthopaedics, including being used as tools for teaching, testing, surgery, and research. Teaching applications allow students to access new digitized anatomic databases and to dissect anatomy in virtual space to observe complex three- dimensional (3-D) anatomic relationships from unlimited visual perspectives, eg, with the Visible Human (Center for Human Simulation, University of Colorado). Testing skills in virtual space is in the early stages of development but it will allow teachers, clinical faculty, and board examiners to teach and to test students, residents, and practicing surgeons on standardized models to evaluate their procedural skills used in the clinic or operating room. In the surgical suite, virtual images are being generated by computer-assisted surgical-navigation machines (these machines are being manufactured by Medtronic Surgical Navigation Technologies, Louisville, Colorado; Stryker Navigation, Kalamazoo, Michigan; BrainLAB GmbH, Munich, Germany; and Aesculap AG, Tuttlingen, Germany) to guide surgeons through complicated hip, knee, and spine procedures with the promise of new algorithms to guide surgeons through trauma surgery in the near future. Research applications have been slower to evolve, but we address the application of virtual imaging to test a scientific hypothesis, illustrating the application of the emerging use of virtual reality in orthopaedic research.

It has been well established in the literature that proper alignment is an important aspect in terms of the success and longevity in total knee arthroplasty (TKA).4,18 The consequences of malalignment include potential for aseptic loosening, ligament instability, patella maltracking, patella fracture, accelerated wear, and decreased range of motion.9,10,13,22 The goal of reconstruction is to restore knee kinematics and to limit wear to promote longevity of the reconstruction.6 The morphologic shape of the distal part of the femur dictates the orientation and kinematics of prosthetic TKA.7 As a result, the morphology of the distal femur is a topic attracting considerable interest in the literature with continued disagreement and controversy.1,3,6,7,14,20

To address these issues, mechanical alignment guides have been developed for use in TKA in an effort to achieve accurate axial alignment. Despite improved technology, however, conventional alignment techniques result in a considerable amount of postoperative axial14,21,22,27 and rotational10,16,29 malalignment. Computer-navigation systems are being introduced to improve the accuracy of component alignment in TKA.2,5,12,15,19,24,26,27 However, the ability of a surgeon to identify anatomic landmarks of the distal femur reliably remains paramount in the successful alignment needed in TKA, whether using conventional techniques or the more contemporary computer navigation systems.

Clinical experience and authors of scientific studies suggest that the reliance of traditional alignment instruments and contemporary computer-assisted navigation on morphologic features of the knee leads to malalignment in a significant number of cases. 10,14,16,21,22,27,29 The problems inherent in the algorithms of traditional instruments, ie, positioning the femoral component relative to femoral morphologic features and the tibial component relative to tibial morphologic features without referencing each other, have been incorporated into the algorithms of contemporary computer-assisted navigation. We surmise there will be substantial interobserver variability in the identification of landmarks about the knee.


Our study was conducted at a scientific exhibit at the American Academy of Orthopedic Surgeons in Washington D.C. in February 2005 after approval from the Colorado Multiple Institutional Review Board (COMIRB). Participants included orthopaedic surgeons, residents, fellows, scientists and exhibitors. There were 55 participants in the study: 33 orthopaedic surgeons, seven residents or fellows, and 15 nonsurgeon scientists or exhibitors. Of the participants, 34 indicated that they were familiar with surgical navigation and three reported familiarity with virtual reality.

Individuals were asked to identify four key anatomic landmarks in the distal femur-the medial and lateral epicondyles as well as points equidistant from the articular surface of the posterior femoral condyles when viewing the distal femur from a medial or lateral perspective. Epicondyle 1 represented the me- dial epicondyle, epicondyle 2 represented the lateral epicondyle, condyle 1 represented the center of the lateral condyle, and condyle 2 represented the center of the medial condyle (Fig 1). The two epicondylar points then were used to generate the epicondylar axis and the two condylar centers were used to generate the cylindrical axis (Fig 1).

Fig 1
Fig 1:
A virtual femur viewed from the medial perspective has been extracted from the Visible Human with the epicondyles joined by the upper line (epicondylar axis) and the condylar centers joined by the lower line (cylindrical axis). The lines are computer generated based on the landmarks selected by the participant, as illustrated in Figure 2. The landmarks are those measured in the study including the epicondyles located where the upper line exits the bone medially and laterally and the condylar centers located where the lower line exits the bone medially and laterally.

The anatomic landmarks were identified on a virtual femur segmented, ie, extracted free of soft tissues, from the Visible Human (University of Colorado Human Simulation Laboratory) and were viewed using 3-D glasses (CrystalEyes, StereoGraphics, Beverly Hills, CA). The four landmarks identified by the observer were reduced to points in space with coordinates in three dimensions (x, y, and z) using a probe (Fig 2) linked to a haptic device (SensAble Technologies, Woburn, MA). The haptic device is an articulated arm (Fig 2) with transducers and motors built into the articulations. The transducers send positional signals to the computer to allow constant tracking of the probe tip and send coupled signals to the motors to provide programmed resistance to the arm allowing the operator to “feel” the surface of the virtual bone. When activated by a foot switch, the computer captured the coordinates generated by the haptic transducers, which corresponded to the landmarks in contact with the probe tip. The position of the femur in space was standardized for comparison between observers by calibrating or registering the tip of the probe relative to the frame holding the haptic before each observer collected data.

Fig 2
Fig 2:
A study participant is shown holding a probe connected to the haptic device. A real femur is located in the approximate position of the virtual femur “observed” by the participant wearing 3-D Crystal Eyes and “touched” with the probe attached to a haptic device. The location of landmarks is recorded by a computer connected to the haptic coupled to the probe.

Data points were analyzed in pairs to establish the relative location of each participant's selected point to each of his/her co-participant's point selection. This method permitted the investigators to assess how closely each participant agreed with each of the other co-participants. This approach avoided imposing an investigator bias, ie, it avoided the need to establish a benchmark or “true” morphologic landmark against which each participant was measured. However, this approach precluded measuring a mean and standard deviation from a single location for all observers.

Participant data points were analyzed by calculating the Euclidean distance between two points at a time (ie, the same point identified by two different observers) for a total of 1485 comparisons of interobserver reliability in the selection of each of the four anatomic landmarks of the distal femur. Comparisons were also made to identify any substantial observed difference between the two landmarks on each side of the femur, eg, between the epicondyles and the condylar center points, for each individual observer.

Comparisons were done using student's t tests and Wilcoxon signed-rank test (a nonparametric t test). The software used was SYSTAT 8.0 (SPSS Inc., Chicago, IL) for the t tests and the graphs. GraphPad Prism 4 (GraphPad Software, Inc., San Diego, CA) was used for the Wilcoxon test.


Each landmark, compared two at a time, was different (p < 0.001) from its comparator with a minimum distance of 0.2 mm (Table 1). Therefore, for each of the six Euclidian distances calculated between points identified by two observers (two points each with x, y, and z coordinates) were not equal to a mean of 0 (Figs 3-6). This observation indicates that no two observers identified the same location for any of the four landmarks measured. The lateral epicondyle demonstrated the least interobserver variability. However, the mean and standard deviation for each landmark (epicondyle and condylar center) were similar, suggesting there is little difference between landmarks from the standpoint of observer variability. In addition, comparisons between averages for the medial epicondyle and medial condylar center points and for the lateral epicondyle and the lateral condylar center points were different (p < 0.001) indicating these two references are not the same, ie, the epicondylar line and the cylindrical line are not collinear (Table 1, Fig 1).

Observer Differences Measured in Millimeters
Fig 3
Fig 3:
This graph of data illustrates the variability encountered by multiple individuals identifying the medial epicondyle (Epicondyle 1). The horizontal axis reflects the distance between the medial epicondyles identified by two separate observers (measured in millimeters).
Fig 4
Fig 4:
This graph of data illustrates the variability encountered by multiple individuals identifying the lateral epicondyle (Epicondyle 2). The horizontal axis reflects the distance between the lateral epicondyles identified by two separate observers (measured in millimeters).
Fig 5
Fig 5:
This graph of data illustrates the variability encountered by multiple individuals identifying the lateral condylar centers (Condyle 1). The horizontal axis reflects the distance between the lateral condylar centers identified by two separate observers (measured in millimeters).
Fig 6
Fig 6:
This graph of data illustrates the variability encountered by multiple individuals identifying the medial condylar centers (Condyle 2). The horizontal axis reflects the distance between the medial condylar centers identified by two separate observers (measured in millimeters).


It has been well established in the literature that proper alignment is an important aspect in terms of the success and longevity in TKA.4,18 Historically, axial alignment, or alignment in the coronal plane, has been emphasized as a critical factor in the longevity of TKA. Several authors have concluded that achieving a postoperative limb axis within ± 3° of varus or valgus from a straight line connecting the hip, knee and ankle, ie, the mechanical axis, is associated with a lower rate of aseptic loosening.4,14,18,23 The mechanical alignment guides used in TKA have evolved through the years in an effort to achieve this axial alignment. Despite improvements in instrumentation, however, conventional alignment techniques result in a considerable amount of postoperative malalignment. 10,14,16,21,22,27,29 To address this important issue, computer algorithms are being written and computer- assisted navigation techniques are being introduced to improve the accuracy of component alignment in TKA. 2,5,12,15,19,24,26,27

In this study, we do not challenge this recognized impact of alignment on outcome in TKA, and we do not challenge the value brought to this task by conventional instrumentation and contemporary computer-assisted surgical navigation. We raise the concern, however, that the reliance of conventional and contemporary alignment algorithms on morphologic features of the knee introduces a source of variability and error in an otherwise accurate procedure. This study tests the hypothesis that observer variability occurs in the selection of morphologic landmarks, using a method of repeated measures by multiple observers on a femur in virtual space.

The repeated measures method used in this study has been advocated8 and embraced by others studying the alignment of the knee.11,17,25,28 There are no authors of studies for comparison, however, who specifically address the relationship measured in this study between the epicondylar axis and the cylindrical axis of the knee. These axes are not physical entities but virtual lines defined by morphologic features of the femur. It is this virtual aspect of the axes which prompted the use in this study of a haptic device to measure an anatomic specimen in virtual space.

The use of a virtual anatomic specimen is unique to this study and we have not found it reported in earlier work published in the literature. In general terms, the advantages for testing with virtual specimens include access to existing normal and pathologic digitized specimens, access to new specimens by digitizing patient knees, access to specimens with and without soft tissues, freedom from the inconvenience of cadavers, and no intraoperative delays imposed on patients to collect data. Digitized specimens from patients also offer the advantage of studying alignment in a variety of pathologic processes, eg, varus and valgus, not easily found in cadaveric material. Added to these general advantages, repeated measures are made easily on the same knee(s) by one or more observers, as illustrated by this study. Despite these potential advantages, no studies done in virtual space are available for comparison. No studies were identified in the published literature that employed a digitized femur, with the possible exception of several studies relying on radiographs that may have been digitized.17,25,28

The clinical significance of this study is yet to be determined because there are no comparable reports in the literature of morphologic measures made on a digitized specimen in virtual space. Nor are there similar morphologic measures reported for a real specimen. Studies are needed to correlate data collected in virtual space with data collected in the clinical setting. A study comparing the selection of landmarks by multiple individuals on a virtual knee and a knee from a cadaver or patient would help to validate the results of this study. Despite the absence of this validation, we think is evident from the results presented here that a statistically significant variation exists between observers when measuring the epicondyles or condylar centers. The median and mean variation documented in this study would translate to approximately 5° of component rotation in a typical TKA, with a range of 0° to 30° of component rotation for either morphologic landmark, ie, either epicondyles or condylar centers.

Implicit in the success of any alignment technique in TKA, whether conventional or computer assisted, is the ability of the surgeon to accurately and consistently identify key anatomic landmarks in the patient's knee. In a patient who has degenerative disease of the knee, the anatomy can become distorted, making it more difficult to identify key anatomic landmarks accurately. It also is difficult to define landmarks covered by soft tissue, eg, the collateral ligament attachments covering the epicondyles. Even in the absence of pathologic deformity and soft tissue obstructions, there is an issue of variability in observer selection of otherwise normal and unobstructed landmarks, as illustrated in this study. With the recent integration of computer navigation systems into total joint replacement surgery, it is important for investigators to identify accessible and reproducible landmarks. Without the accurate input of reference points corresponding to morphologic features of the knee, the result obtained with computer-assisted navigation may be implants that are malaligned more precisely.

It has been well established that proper alignment is essential to the success and longevity of TKA. Several anatomic landmarks of the distal femur are used by orthopaedic surgeons to help guide component positioning to achieve this improved alignment and its associated longevity. Through our results, we show that certain anatomic landmarks used in TKA are not reliable when tested in virtual reality. This use of virtual reality to demonstrate variability in landmark selection represents one of many potential scientific and clinical applications for orthopaedic research, and it may be useful in the future for other studies to improve the success of TKA.


The authors thank Mary Samson for her assistance in the statistical analysis of data and for the preparation of charts and tables for this manuscript, David Rubenstein for his original computer software, Todd Baldini for his custom hardware, and Vic Spitzer for his Visible Human data.


1. Arima J, Whiteside LA, McCarthy DS, White SE. Femoral rotational alignment, based on the anteroposterior axis, in total knee arthroplasty in a valgus knee. J Bone Joint Surg. 1995;77A:1331-1334.
2. Bathis H, Perlick L, Tingart M, Lüring C, Zurakowski D, Grifka J. Alignment in total knee arthroplasty. A comparison of computer- assisted surgery with the conventional technique. J Bone Joint Surg. 2004;86B:682-687.
3. Berger RA, Rubash HE, Seel MJ, Thompson WH, Crossett LS. Determining the rotational alignment of the femoral component in total knee arthroplasty using the epicondylar axis. Clin Orthop Relat Res. 1993;286:40-47.
4. Coventry MB. Two-part total knee arthroplasty. Evolution and present status. Clin Orthop Relat Res. 1973;145:29-36.
5. Decking R, Markmann Y, Fuchs J, Puhl W, Scharf H. Leg axis after computer-navigated total knee arthroplasty. J Arthroplasty. 2005;20:282-288.
6. Eckhoff DG, Bach JM, Spitzer VM, Reinig KD, Bagur MM, Baldini TH, Rubinstein D, Humphries S. Three-dimensional morphology of the distal part of the femur viewed in virtual reality, Part II. J Bone Joint Surg. 2003;85A (Suppl 4): 97-104.
7. Eckhoff DG, Dwyer TF, Bach JM, Spitzer VM, Reinig KD. Three- dimensional morphology of the distal part of the femur viewed in virtual reality, Part I. J Bone Joint Surg. 2001;83A (Suppl 2): 43-50.
8. Eliasziw M, Young SL, Woodbuty MG, Fryday-Field K. Statistical methodology for the concurrent assessment of interrater and intrarater reliability: using goniometric measurements as an example. Phys Ther. 1994;74:777-788.
9. Fehring TK, Odum S, Griffin WL, Mason JB, Nadaud M. Early failures in total knee arthroplasty. Clin Orthop Relat Res. 2001;392: 315-318.
10. Figgie HE, Goldberg VM, Figgie MP, Inglis AE, Kelly M, Sobel M. The effect of alignment of the implant on fractures of the patella after condylar total knee arthroplasty. J Bone Joint Surg. 1989;71A: 1031-1039.
11. Green CC, Edwards TB, Wade MR, Carson EW. Reliability of the quadriceps angle measurement. Am J Knee Surg. 2001;14:97-103.
12. Haaker RG,Stockheim M. Kamp M, Proff G, Breitenfelder J, Ottersbach A. Computer-assisted navigation increases precision of component placement in total knee arthroplasty. Clin Orthop Relat Res. 2005;433:152-159.
13. Hungerford DS, Krackow K, Kenna RV. Total Knee Arthroplasty: A Comprehensive Approach. Baltimore: Williams & Wilkins; 1984.
14. Jeffery RS, Morris RW, Denham RA. Coronal alignment after total knee replacement. J Bone Joint Surg. 1991;73B:709-714.
15. Jenny JY, Boeri C. Navigated implantation of total knee endopros- thesis, a comparative study with conventional instrumentation. Z Orthop Ihre Grenzgeb. 2001;139:117-119.
16. Lewis P, Rorabeck CH, Bourne RB, Devane P. Postero-medial tibial polyethylene failure in total knee replacements. Clin Orthop Relat Res. 1994;299:11-17.
17. Lonner JH, Laird MT, Stuchin SA. Effect of rotation and knee flexion on radiographic alignment in total knee arthroplasties. Clin Orthop Relat Res. 1996;331:102-106.
18. Lotke PA, Ecker ML. Influence of positioning of prosthesis in total knee replacement. J Bone Joint Surg. 1977;59A:77-79.
19. Mielke RK, Clemens U, Jens JH, Kershally S. Navigation in knee endoprosthesis implantation, preliminary experiences and prospective comparative study with conventional implantation technique. Z Orthop Ihre Grenzgeb. 2001;139:109-116.
20. Miller MC, Berger RA, Petrella AJ, Karmas A, Rubash HE. Optimizing femoral component rotation in total knee arthroplasty. Clin Orthop Relat Res. 2001;392:38-45.
21. Peterson TL, Engh GA. Radiographic assessment of knee alignment after total knee arthroplasty. J Arthroplasty. 1988;3:67-72.
22. Rand JA, Coventry MB. Ten-year evaluation of geometric total knee arthroplasty. Clin Orthop Relat Res. 1988;232:168-173.
23. Robertsson O, Knutson K, Lewold S, Lidgren L. The Swedish Knee Arthroplasty Register 1975-1997: an update with special emphasis on 41,223 knees operated on in 1988-1997. Acta Orthop Scand. 2001;72:503-513.
24. Sikorski JM, Chauhan S. Computer-assisted orthopaedic surgery: Do we need CAOS? J Bone Joint Surg. 2003;85B:319-323.
25. Schmidt GL, Altman GT, Dougherty JT, DeMeo PJ. Reproducibility and reliability of the anatomic axis of the lower extremity. J Knee Surg. 2004;17:141-143.
26. Sparmann M, Wolke B, Czupalla H, Banzer D, Zink A. Positioning of total knee arthroplasty with and without navigation support, a prospective, randomized study. J Bone Joint Surg. 2003;85B:830-835.
27. Stulberg SD. How accurate is current TKR instrumentation? Clin Orthop Relat Res. 2003;416:177-184.
28. Takahashi T, Yamanaka N, Komatsu M, Ogawa Y, Yoshida S, Yamamoto H. A new computer-assisted method for measuring the tibio-femoral angle in patients with osteoarthritis of the knee. Osteoarthritis Cartilage. 2004;12:256-259.
29. Wasielewski RC, Galante JO, Leighty RM, Natarajan RN, Rosenberg AG. Wear patterns on retrieved polyethylene tibial inserts and their relationship to technical considerations during total knee arthroplasty. Clin Orthop Relat Res. 1994;299:31-43.
© 2006 Lippincott Williams & Wilkins, Inc.