Can tibial coverage in total knee replacement be reliably evaluated with three-dimensional image-based digital templating?

Article focus

We aimed to assess the intra- and interobserver reliability of a three-dimensional digital templating method based on CT reconstruction for evaluating tibial coverage in total knee replacement (TKR)

It was hypothesised that the method was reliable, regardless of the reference axis used to align the tibial implant

Key messages

Contemporary methods for TKR increasingly rely on digital technologies for pre-operative planning. However the reliability of these methods with respect to tibial coverage remains unknown

This study addresses this gap in the current knowledge and describes a robust method for templating tibial coverage using CT-derived models of tibiae and various tibial components

Further work is required to explore the optimisation between tibial coverage, particularly of the cortical shell, and component rotation when aligning the component to the resected bone surface

Strengths and limitations

The results are founded on robust methods of templating with immediate clinical utility and a detailed statistical analysis with respect to intra- and interobserver reliability

Further work with a larger sample size is required to generalise to the wider population and with the vast variety of tibial components available

The definition of ‘coverage’ in the present context requires more precise delineation for the assessment of the efficacy of tibial components

Introduction

Interest in templating based on three-dimensional (3D) imaging has recently emerged in parallel with the growing enthusiasm for patient-specific instrumentation. The accurate placement of components on the tibia is important to achieve maximum coverage of the resected plateau, thereby imparting good initial implant stability,^1-3 enhancing load transfer across the proximal tibia^1,4-6 and decreasing the potential for implant subsidence.^1,2,7

In order to optimise patellofemoral tracking, a number of studies have reported rotation with reference to the medial margin of the anterior tibial tuberosity (ATT)⁸ or to the posterior condylar axis of the tibia.⁹ Many other axes of reference have been described, such as the mid-sulcus of the tibial spine,¹⁰ the distal reference axes of the foot or ankle^9,11,12 or the femoral transepicondylar axis (TEA).^13,14 However, the literature suffers from a lack of consensus regarding the effect of rotational alignment on coverage.¹⁵ The controversy could be related to differences in methods, but importantly the reliability of the templating and measurement method employed is consistently not reported.^7,16-18

Various templating methods have been used to estimate the theoretical tibial coverage afforded by different prosthesis designs.^7,16-18 Two-dimensional (2D) tibial models created from photographed resection specimens obtained during TKR, from cadavers¹⁷ or digitised unmagnified radiographs of retrieved surgical specimens, have been described.^2,16,19 However, the reliability of these 2D methods has never been reported, and could be questioned due to the lack of perspective and the obstruction of key anatomical landmarks when fitting the component template.

Modelling of the tibial structure in 3D has been described in cadaveric specimens,⁷ but surprisingly no recent publications have focused on digital templating of resected tibiae with contemporary modelling methods. We describe such a method based on low-dose CT imaging combined with 3D modelling. As with every new method, demonstrated reliability is a fundamental parameter before application to the clinical setting, in contrast to previous efforts.

In order to address the gap in the current literature, the purpose of this study was to assess the intra- and interobserver reliability of a 3D digital templating method based on CT reconstruction for evaluating tibial coverage in TKR. Our hypothesis was that the method was reliable regardless of the reference axis used to align the tibial implant.

Materials and Methods

Tibial model preparation

After import into SolidWorks, each model was digitally resected 10 mm below the articular surface, using the lowest point on the lateral tibial plateau as a reference point. The resection was made perpendicular to the long axis of the tibia in order to simulate the tibial cut made intra-operatively during TKR (Fig. 1). Following resection, each tibial plateau was digitally marked in the axial view with two axes of rotation for reference during templating (Fig. 2). The anterior tibial tubercle axis (ATT) was defined by a line connecting the centre of the posterior cruciate attachment on the tibia and a point located at the medial third of the anterior tibial tubercle. The two points were determined for each tibia by mutual agreement between three orthopaedic surgeons (AJC, SL and JCB). The posterior condylar axis (PCA) was defined by a line drawn perpendicular to the middle of a second line connecting the most posterior points on the medial and lateral tibial condyles respectively.

Fig. 2

Diagram showing the anterior tibial tubercle axis (ATT; solid grey line) and the posterior condylar axis (PCA; dashed line).

Fig. 1

Three-dimensional CT reconstructions before (left) and after resection (right).

Each tibia was also marked to define sub-regions of interest including the anterior, posteromedial and posterolateral regions (Fig. 3). To define these sub-regions, a medial to lateral line was drawn perpendicular to the posterior condylar axis line. This line was then shifted in the anterior-posterior direction until the widest medial to lateral portion of the tibia was located, which was termed the ML line. From the midpoint of the ML line, two additional lines were drawn at 60° in the posteromedial (PM line) and posterolateral directions (PL line).

Fig. 3

Diagram showing the sub-regions of interest (PM, posteromedial line; PL, posterolateral line, ML, mediolateral line; PCA, posterior condylar axis).

Figs. 4a - 4b

Diagrams showing the tibia template, with a) the asymmetric prosthesis aligned with the surgical axis, with no overhang permitted, and b) the symmetric prosthesis aligned with the surgical axis, with minimal posterolateral overhang permitted (*).

Results

In total, each surgeon completed 120 template procedures in accordance with the criteria and conditions described.

Excellent intra-observer reliability (ICC ≥ 0.94; TEM ≤ 18.1 mm²) was observed when the trays were aligned to the ATT axis in the NO condition (Table II). Of the sub-regions, the posteromedial region displayed lower TEM (3.4 mm² to 8.8 mm²) compared with the anterior region (15.1 mm² to 18.1 mm²). A similar pattern was observed for the OV condition (Table II), with excellent reliability overall (ICC ≥ 0.95; TEM ≤ 17.2 mm²) and differences in TEM between anterior (12 mm² to 17.2 mm²) and posteromedial (2.9 mm² to 6.6 mm²) sub-regions.

When the trays were aligned to the PCA, excellent intra-observer reliability was observed for total and sub-region coverage in the NO (ICC ≥ 0.76; TEM ≤ 22.4 mm²) and OV (ICC ≥ 0.85; TEM ≤ 25.7 mm²) conditions. Intra-observer reliability of total coverage was comparable with the ATT condition, although ICCs were lower and TEM higher for sub-region coverage in the PCA condition (Table III). Nevertheless, a similar pattern was observed across both rotation-coverage combinations, with differences in TEM between anterior and posteromedial sub-regions (Table III).

Intra-observer reliability of total coverage using the BFA to set rotation was excellent in the NO condition (ICC 1.0; TEM ≤ 12.1 mm²) (Table IV). Similarly, sub-region coverage reliability was good to excellent (ICC 0.58 to 0.98; TEM ≤ 25.7 mm²). In the OV condition, total coverage reliability was comparable (ICC 0.73 to 1.0), although TEM was larger than for other rotation-overhang combinations (≤ 64.7 mm²). A similar pattern was observed for sub-region reliability with ICC ranging from poor (0.25) to excellent (0.95) and higher TEM than for other conditions (≤ 45.6 mm²) (Table IV). Of note for the BFA condition was comparable reliability for the anterior and posteromedial sub-regions.

Interobserver reliability of coverage was superior when trays were aligned using the ATT axis in the NO condition (ICC 0.94 to 1.0; TEM 8.3 mm² to 17.5 mm²) when compared with PCA (ICC 0.72 to 0.94; TEM 17 mm² to 54.5 mm²) and BFA (ICC 0.48 to 0.98; TEM 23.3 mm² to 47.1 mm²) (Table V). In contrast, this pattern was not replicated in the OV condition (Table V), with ICC ranging from good (0.6) to excellent (1.0) for all axes and ATT total coverage displaying the highest TEM (56.2 mm²) of any axis.

Discussion

The main finding of this study was that excellent reliability of tibial coverage can be obtained using 3D CT reconstruction and digital templating. While aligning components with the ATT axis produced the most reliable results for sub-region coverage, other axes demonstrated more variability in these measurements. Increased variability was particularly evident when no rotational control was used. Measurement of tibial coverage is overall reliable using the templating method described even if the rotational axis selected still has a minor influence.To our knowledge, this is the first study to report on the reliability of a templating and measurement method for assessing tibial coverage in TKR.

Even if digital templating could potentially revolutionise the conduct of surgical planning or sizing in the future, caution should be applied when interpreting the results if the reliability of the measurement is not well documented.^7,16-18 Moreover, there are several disadvantages to previously reported methods, being time consuming and labour intensive, often requiring cadavers or surgical resection specimens and being prone to magnification or other measurement errors.^2,7,16-18 A previous study compared eight different tibial tray designs, including six symmetric and two asymmetric designs.¹⁷ However, limited details were given regarding the criteria for choosing the template size and position to calculate tibial coverage. Furthermore, no attempts were made to control for component rotation and the authors did not quantify the reliability of the technique. Others have reported differences in coverage between template designs despite repeating these methodological deficiencies. Westrich et al¹⁶ failed to describe the criteria for positioning the templates in the medial-lateral or anterior-posterior direction. Similarly, no attempts were made to determine the reliability of the templating and measurement technique. In another study,⁷ 3D tibial models were generated from cadaveric specimens and templated digitally with tibial trays from one symmetric and one asymmetric prosthesis design. The criteria used to position the templates in this study were defined, in contrast to previous studies, however all templating was carried out by a single surgeon and no attempt was made to assess the reliability of the templating and measurement technique. As a result of these limitations, there remains no gold standard method to assess tibial coverage, making it difficult to compare the results. Potentially, the present study could serve this purpose for future investigations in this area.

There are several advantages to the templating and measurement method we describe compared with earlier efforts. Firstly, 3D digital tibial models were generated from CT scans. The use of CT scans over cadavers has numerous advantages in that the potential sample size is increased and demographic data more readily available.¹⁵ Furthermore, computer digitisation of data points eliminates magnification errors and allows for manipulation in space to view the knee in the ideal orientation. A recent study reported excellent reliability using 3D-CT reconstruction to measure the position and orientation of TKR components post-operatively,²² thereby presenting a potential vehicle for future validation of the templating and measurement method described here. Further, digital templating each tibia in the present study facilitated efficiency in templating, standardised coverage calculations and ensured rotational and over-hang criteria were strictly followed. Lastly, as digital templates are now readily available from manufacturers, design changes and newer prostheses can be easily evaluated and comparisons can be made to previously tested prostheses under identical experimental conditions. In this way, labour, research costs and potential errors associated with replicating previous manual measurements, or cadaveric studies can be reduced.

Despite the advantages of our reported technique, there are also limitations of the study to note, aside from access to software to perform the templating and calculations, which may be prohibitive for some groups. First, the small sample of five tibial models is not likely to be representative of the wide variation in tibial morphology encountered by the surgeon during TKR.²³ However, templating a larger number of tibiae using the described technique is unlikely to improve the reliability reported here, although the findings would be more generalisable. Secondly, two prosthesis designs were tested, one symmetric and one asymmetric. Templating for different prosthesis designs may be less reliable, particularly if a larger spectrum of tibial models is tested. Thirdly, the concept of ‘coverage’ itself is not well-defined and its clinical application is hampered by a lack of consensus. Some have emphasised covering the entire resected surface,^4,17 others have focused on optimising coverage in specific regions^2,24-26 and still others have stressed coverage of the cortical shell.^3,27 We demonstrated the satisfactory reliability of our method for total and sub-region surface of the total resected surface, but further studies will be required to assess cortical shell coverage. It is also important to note that we tested only two fixed rotational axes, the ATT and the PCA. There is currently no consensus with regard to the rotational alignment of the tibial component and the design might restrict the choice of rotational axis.¹⁵ The reliability of templating and measurement techniques may therefore vary considerably if other rotational axes are used.