Computer navigated total knee arthroplasty: Aspects of a single unit's experience of 777 cases

Abstract

The use of computer navigation and conventional techniques in total knee arthroplasty remains controversial. Advocates of computer navigated techniques cite better alignment of components and reduced morbidity associated with avoidance of intramedullary instrumentation as a rationale for their use. In contrast, proponents of conventional techniques argue that better alignment does not correlate with a better functional outcome and that the conventional approach avoids the perceived risk of fracture associated with bicortical insertion of navigation tracker pins.

All total knee arthoplasties performed at our institution are prospectively monitored for life in a dedicated Joint Replacement Assessment Clinic (JRAC). Patients are reviewed by physiotherapists, independent of the surgeons who performed surgery, both preoperatively and at six weeks, three and six months, and one, two and five years postoperatively (and every five years thereafter). Patients are assessed using validated outcome measures (Knee Society Score, Western Ontario and McMaster Universities (WOMAC) osteoarthritis index, Short Form SF-36 Health Survey (version 2) and a patient satisfaction score). In addition, at 6 months post surgery, a CT scan of each implanted prosthesis is performed using the Perth CT knee protocol.

The findings of a single unit's experience of 777 navigated primary total knee replacements are discussed and critically compared to the body of literature that currently relates to this controversial topic.

• Computer navigated total knee arthroplasty
• Perth CT knee protocol

Introduction

Computer navigation for total knee arthroplasty (TKA) is becoming increasingly popular, although significant geographical variations exist both regionally and internationally in the use of this technology. First introduced in 2001, Australian Orthopaedic Association National Joint Replacement Registry data shows that in 2008 11.1% of all primary TKAs performed in Western Australia used computer navigation (this compares with 6.1% in New South Wales and 3.4% in the Northern Territory) [1], whilst United Kingdom Joint Registry data shows that 2.0% of all TKAs performed in 2008 used navigation technology [2]. Computer navigated TKA now represents the default technique in our unit.

To date we have performed 777 computer navigated TKAs. All TKAs performed at our institution are prospectively monitored for life in our dedicated Joint Replacement Assessment Clinic (JRAC). Located in the most isolated city in the world, emigration rates in our TKA demographic are exceptionally low, meaning that very few patients are lost to follow-up. Patients are reviewed by physiotherapists, independent of the surgeons who performed the surgery, both preoperatively and at six weeks, three and six months, and one, two and five years postoperatively (and every five years thereafter). Patients are assessed using validated outcome measures (Knee Society Score [3], [4], Western Ontario and McMaster Universities (WOMAC) osteoarthritis index [5], Short Form SF-36 Health Survey (version 2) [6], [7] and a patient satisfaction score). In addition, at 6 months post surgery, a CT scan is performed for each TKA according to the Perth CT knee protocol [8]. This is a cadaver-based, validated tool and allows the accurate determination of component alignment in three dimensions (coronal varus/valgus alignment, sagittal flexion/extension, and axial external/internal rotation for both femoral and tibial components). This creates an enormous reservoir of data upon which to base an informed commentary on many aspects of computer navigated TKA.

Alignment versus function

Outcome in TKA is multifactorial; however, the restoration of a patient's normal mechanical axis is thought to be of paramount importance. It is now generally accepted that computer navigation produces better alignment than conventional jig-based techniques [9–12]. Despite this, controversy still exists with respect to this statement. In a 2007 meta-analysis of 33 studies comparing navigated with conventional TKA (including 11 randomized controlled trials), Bauwens et al. [13] failed to conclude that navigation offered significant benefit in terms of component alignment when compared with conventional techniques. In parallel, Mason et al. [12], also in 2007, published a meta-analysis of alignment outcomes in navigated TKA citing 29 studies (including 9 randomized controlled trials), all of which were also included in the meta-analysis of Bauwens et al. Mason et al. concluded that a significant improvement in component orientation and a better restoration of the mechanical limb alignment was achieved with computer navigation in comparison with conventional techniques, and they were critical of Bauwens et al.'s methodology and analyses.

In 2004 we published the results of a prospective randomized study comparing component alignment obtained in computer navigated and conventional TKA [14]. A cohort of 71 patients undergoing Duracon (Stryker Orthopaedics, St. Leonards, Australia) TKA without patellar resurfacing was prospectively and randomly allocated to undergo operation using either a computer navigated technique (35 patients) or a conventional jig-based method (36 patients). The Stryker image-free computer navigation system (version 1.0; Stryker Orthopaedics) was used for navigated TKA. Conventional TKA was performed according to the manufacturer's manual, with intramedullary instrumentation for the femur and extramedullary instrumentation for the tibia. All patients underwent postoperative CT scanning according to the Perth CT protocol [8] to assess component alignment. A significant improvement in alignment of components was found in the computer navigated group compared with the conventional group.

In historical series of conventional TKA, loosening, instability and early component failure are recognized consequences of malalignment. However, the definition of “malalignment” remains controversial, and contradictory studies do exist. Rand et al. [15], in 193 TKAs at a mean of 11 years post surgery, found increased aseptic loosening in components with more than 3° of varus deformity, and Ritter et al. [16] reported 8 cases of aseptic loosening in 421 TKAs, five of which were in varus malalignment (overall, however, 235 of the TKAs were in varus malalignment). In contrast, Jonsson et al. [17], in 86 TKAs at a mean of 6.5 years post surgery, reported 5 cases of aseptic loosening in normally aligned TKAs and good function in all knees aligned from 6° varus to 7° valgus. Similarly, in 126 knees at a mean of 19 years post surgery, Ma et al. [18] reported aseptic loosening in 5 cases, all with normal alignment, and good function in 64 knees that were available for review aligned from 3° varus to 15° valgus.

Almost universally, the acceptably quoted “standard” for component alignment in the coronal plane is ±3° to the mechanical axis. Although perpetuated, the source of this standard is infrequently cited. It was initially quoted by Jeffery et al. in 1991 [19], and relates to the outcome of 115 Denham total knee replacements implanted into a heterogonous cohort of patients with osteo- and rheumatoid arthritis. The Denham prosthesis comprises a tibial component with a removable intramedullary guide and a stemmed femoral component (a prosthesis described by Sikorski as “a very unusual knee design” [20]). Furthermore, 3° is an arbitrary figure, and there is no reason to believe that it represents a definitive value for the acceptability of alignment [20]. There is unlikely to be such a universal standard for all TKAs, that is, the standard will vary according to the implant and is likely to be broader as prostheses become increasingly “engineered” (which in turn may explain, in part, the variability in the aforementioned studies). Despite this, ±3° continues to be applied in the research scenario, when in fact it probably more accurately represents the reproducibility that the reasonable arthoplasty surgeon should be able to achieve by conventional means.

A second issue is the understanding that component alignment is not limited to the two-dimensional coronal plane, that is, it is a three-dimensional (3D) phenomenon. The femoral component may be malaligned in the coronal (varus/valgus), axial (external/internal rotation) and/or sagittal (flexion/extension) planes, and component malalignment in one plane, e.g., femoral varus/valgus, may be compensated for by alignment in another plane, e.g., femoral external/internal rotation. Thus, assessment of alignment in one plane only may neglect significant variability in a second or third plane which may influence overall outcome. (Refer to Ritter et al. [16], who reported eight cases of aseptic loosening in 421 TKAs, of which 235 were in varus malalignment. Five of the eight cases of aseptic loosening were in varus malalignment, implying that 230 knees in varus malalignment did NOT loosen.) The use of a malalignment index summing alignment measurements in three planes has been introduced by Sikorski [21]. The 3D assessment of alignment is now easily performed with tools such as the Perth CT knee protocol, and this should represent the gold standard in such studies.

But does an improvement in alignment result in improved functional outcome? Whilst a plethora of literature exists pertaining to component alignment in computer navigated TKA, very few studies have correlated such findings with functional outcome [10], [11], [13], [22]. We previously reported two-year results showing that no functional improvement in outcome was achieved with navigated TKA in comparison with conventional TKA, despite improved component alignment [22]. These results have now been extended to a minimum of five years’ follow-up. No statistically significant difference was seen between cohorts for any component of the Knee Society, WOMAC or SF-36 scores or for patient satisfaction. Similarly, within each cohort, no statistically significant difference was seen in longitudinal data between two and five years for any of the aforementioned outcome score components, showing that function was well maintained [23].

Given this lack of significance in all statistical analyses, we questioned whether these findings represented a true result between groups or if our study simply lacked power. Between two and five years, 18 patients from our initial cohort of 71 patients had died. To address this, a larger group of patients was retrospectively reviewed and used for comparison. During the same period as our randomized controlled study, 100 navigated and 70 conventional total knee replacements were undertaken at our institution. All were reviewed in an identical manner to our study patients, all of whom completed outcome reviews to five years. The results in the larger retrospective cohort showed no statistically significant difference in any measure of functional outcome. In addition, functional scores obtained from review of this larger retrospective cohort were no different to those obtained in the prospective randomized study. This clearly validated our prospective study results implying that TKA component alignment was not as critical to good functional outcome as previously thought.

Mortality and morbidity

Advocates of conventional TKA cite the increased operative and tourniquet times [24–26] associated with navigated TKA, whilst proponents of navigated TKA have yet to identify a significant clinical advantage over conventional TKA that justifies the significant hardware costs perceived to be associated with using such techniques [27]. Studies do exist which show reduced blood loss [14], [28], [29], fewer systemic emboli [24], [30], [31] and fewer clinically detectable thromboembolic events [32] in navigated TKA compared with conventional TKA, and this is thought to be directly related to the avoidance of femoral and/or tibial intramedullary instrumentation. Cohorts in all these studies were, however, very small and offer low levels of evidence, and contradictory evidence does exist showing no difference in blood loss [25], [33] and thromboembolic phenomena [13] between the two groups. More recently, in a prospective study examining cognitive change following TKA, Haytmanek et al. [34] found no difference in postoperative cognitive function between navigated and conventional TKA cohorts. In a commentary on this methodologically well-designed study, Masri recommended that “a larger sample size may be required to definitively prove that there is no such difference” [35] (Haytmanek et al. having only reviewed 61 patients in total).

We recently performed a retrospective review of 327 patients (146 patients undergoing navigated TKA and 181 undergoing conventional TKA) to establish whether navigation, with its avoidance of intramedullary instrumentation, resulted in lower early postoperative morbidity when compared with conventional techniques. Cohorts were well matched in terms of age, gender, body mass index and American Society of Anesthesiologists grade. Statistically significantly longer operative times (8 min) and tourniquet times (8 min) were seen in the navigated cohort. No statistically significant differences were seen in any measure of postoperative morbidity, including length of hospital stay, hemoglobin drop, transfusion rate, incidence of postoperative anemia, time to achieve knee flexion of 70°, and thromboembolic phenomena as measured by the incidence of postoperative pyrexia and acute confusional state. We therefore concluded that navigated TKA with its avoidance of intramedullary instrumentation offered no benefit in terms of early postoperative morbidity when compared with conventional jig-based techniques [36].

Navigation systems

Differing navigation systems now exist, and all should undergo clinical validation prior to widespread application. In our unit we use the Stryker full navigation system and, to a lesser extent, the Stryker Articular Surface Mounted (ASM) system. We have validated the former system as part of a prospective randomized study, finding that it produced statistically significantly better 3D component alignment compared with conventional methods [14]. More recently, in a further prospective randomized study using the Stryker Triathlon Total Knee implant, we showed that the 3D alignment produced using the ASM system showed no statistically significant differences in any of the 12 parameters of the Perth CT knee protocol. In addition, while functional outcome was identical at one year (using the Knee Society Score), operative time using the ASM system was a mean of 10 min less than with the full navigation system. This was statistically significant (p = 0.001) [37].

Increased operative time and the capital cost of navigation systems are two factors frequently cited by advocates of more traditional conventional techniques. The relative importance of these factors needs to be reviewed in perspective. Three recent studies have commented on the increased operative time between navigated and conventional techniques. A recent meta-analysis found that navigation lengthened operative time by a mean of 17 min (range: 14 to 20 min) [13]. Similarly, Dattani et al. quoted a range of 15 to 17 min [38], whilst Lüring et al. reported a mean difference of 20 min (range: 15 to 40 min) [39]. Whilst a significant increase in operative time is unwelcome, the magnitude of this increase is more likely to represent the publishing unit's position on the “navigated TKA” learning curve. In a recent retrospective review of 327 primary TKAs performed in our unit, operative time for navigated TKA was a mean of 8 min longer than for conventional TKA. This value, while statistically significant, was interestingly less than any other mean difference reported in the literature. This clearly reflects the sample sizes being studied, and in turn our unit's position further along the “navigated TKA” learning curve.

The cost of using navigation technology will vary geographically and also between differing healthcare systems. Dong and Buxton produced the most commonly cited assessment of the cost-effectiveness of navigation technology [29], summarizing that (in 2006) navigated TKA costs on average US$430 per case more than conventional TKA, and that this conveyed a small gain of 0.0148 quality-adjusted life years (QALYs) over ten years. This was regarded as being cost-effective, but was based on key assumptions (that 70.4% of complications after TKA are due to improper component positioning and that navigation reduces malalignment rates by 48%) which will result in a cumulative 1.6% reduction in complex revision rates at 10 years. Given the assumptions made, together with more recent functional outcome studies, the forecasted theoretical cost-benefit of navigated TKA must be interpreted with caution.

In the clinical setting the capital cost of a computer navigation system is approximately AUS$250,000. The only additional costs per case are for disposable tracker pins and the sterilization of reusable trackers and pointers. Whilst the capital cost may seem prohibitive, most companies (at least in the public/non-private sectors) will provide these systems on permanent loan and local cost-analyses should allow for this.

In addition to operative time and cost, the risk of fracture though the site of placement of bicortical tracker pins is also cited by some advocates of conventional techniques as a cause for concern when considering the use of navigation technology [35], [36]. The use of the Stryker ASM system removes this risk completely as it uses no bicortical pins; however, the true incidence of such fractures is very low. In 777 cases performed in our unit, we have seen only one such fracture (0.13%). This occurred in an 82-year old female who sustained a mechanical fall directly onto the operated TKA three weeks after surgery. Intraoperatively during TKA, she was found to have very soft bone negating the ability to place stable/non-mobile tracker pins, so navigation was aborted and a conventional TKA performed (Figure 1). Following intramedullary nailing of this injury, the patient fell once again, resulting in a subcapital fracture neck of femur. No loosening of the TKA was seen.

Figure 1. (a) Lateral radiograph post TKA. Cortical defects produced by a femoral tracker pin are clearly visible. (b) Fracture through the femoral pin site after a fall.

Other issues

The use of navigation has a definitive learning curve. Anecdotally, we feel that 30 to 40 cases are required before becoming proficient. The main difficulty associated with navigation systems is registration and modeling of the patient's anatomy. Final alignment in terms of 3D alignment is entirely dependent upon the accuracy of this process, and clearly “if rubbish is put in, rubbish will come out”. Experience is needed with this process, and we have noted that difficulty is most commonly encountered with the registration of the medial and lateral epicondyles (particularly the former). Any surgeon undertaking navigated TKA should also be competent in conventional TKA. Conversion to a conventional technique is warranted if there is any post-registration movement of tracker pins, if concerns exist with the overall registration values obtained, or if a computer malfunction occurs.

In addition, a detailed understanding of knee kinematics is required, together with the ability to interpret changes in kinematics in real time. In the absence of such an understanding, the computer output from a navigation system may seem overwhelming. However, these requirements can often be simplified according to the system (e.g., the ASM) or the specific parameters prioritized by the operating surgeon. Once a working knowledge of knee kinematics is gained, navigation systems are exceptionally user-friendly and facilitate an understanding of the subtleties of knee kinematics that is simply not available when using conventional techniques. This also makes navigation systems a valuable teaching instrument.

The future of navigation

New technologies now exist which may prove superior to navigated TKA. The “personalized” knee is a new concept which uses the patient's own anatomy via imaging to manufacture patient-specific jigs which, when applied to the patient's femur and tibia, identify exactly the position of all cutting blocks. It is effectively a form of “preoperative navigation”. Examples include the Smith & Nephew Visionaire TKA (London, UK) (based on the Genesis II TKA) and the Stryker ShapeMatching TKA (Newbury, UK) (based on the Triathlon TKA). The Biomet Signature Knee (Bridgend, UK) is another such system (based on the Vanguard TKA) and uses a preoperative MRI scan to model exactly the patient's pathological knee contours. An initial review was conducted in which 25 Signature TKAs performed at our institution were assessed using the Perth CT knee protocol. All knees performed by a single surgeon (P.H.) showed that overall alignment was better than that achieved with conventional techniques, but not as good as that achieved using computer navigation techniques. Given that implant companies have moved on to “new technologies”, it appears by definition that they feel unable (at least in the short to medium term) to market navigation technology so as to achieve sustainable revenues. In addition, given that the alignment achieved with the new personalized systems appears to be inferior to that achieved with navigation, it is evident that the latter will not be marketed on the basis of achieving improved alignment or outcome. Despite the inconvenience of requiring a preoperative MRI scan, personalized knee systems do offer one significant advantage compared with navigation and conventional systems: Intraoperatively, only a single tray of implant trials and equipment is required, as component sizes are known prior to surgery. This greatly reduces inventory volumes and also the load on sterilization services. It is likely, therefore, that personalized systems will be marketed on the basis of cost, an increasing area of importance given the finite resources and the constraints that many surgeons’ healthcare systems now advocate.

Conclusion

Despite finding superior and more reproducible three-dimensional component alignment with navigated TKA, in a series of 777 cases we have identified no single factor, in terms of clinical outcome, upon which to base its selective use over conventional techniques. We have found no difference in functional outcome at a minimum of five years' follow-up, and no difference in mortality and morbidity between techniques. Conversely, we feel the lengthened operative time, cost and the associated risk of fracture are invalid variables upon which to support the non-use of navigation technology. In addition, we have validated different navigation systems and regard the technology as a valuable adjunct to surgeon training.

Despite all these factors, our unit continues to use computer navigation as its default technique in TKA. This may seem contradictory, given the conclusions drawn above; however, the rationale for its continued use is also drawn from work undertaken in our department. It is widely appreciated that outcome after TKA is multifactorial and that alignment is only one of many possible outcome determinants. In addition, the measure of outcome that a researcher cites as being relevant may also influence practice.

In a review of 159 patients undergoing TKA at our institution, we found that patients with good component alignment rehabilitated faster and had a significantly shorter postoperative hospital stay (2 days shorter, p = 0.001). Patients with good coronal femoral alignment also had better function at 1 year (p = 0.013) [40].

Evidence as to the multifactorial nature of outcome after TKA was given by a review of “stiff knees” at our institution requiring manipulation after TKA. Component alignment in 21 stiff knees requiring manipulation post TKA and 260 TKAs with a good outcome were compared, and no differences in any measure of alignment were found [41]. Yet further evidence of the multifactorial nature of outcome is found by analyzing component alignment in TKAs with a good functional outcome. When looking specifically at femoral alignment in the coronal and sagittal planes, femoral rotation, and tibial alignment in the coronal and sagittal planes, of 260 TKAs with a good functional outcome 16%, 35%, 34%, 8% and 32%, respectively, were malaligned in the aforementioned planes (after Sikorski [20]). Malalignment was reduced to 7%, 24%, 24% and 4%, respectively, in the first four of the above planes (with no value for sagittal tibial alignment) if Jeffery's standards were applied in three dimensions [19]. Thus, depending on the standards applied, “malaligned” components can still result in a good functional outcome after TKA, therefore other factors must play a significant role.

Having an evidence-based insight into the factors governing outcome after TKA, navigation continues to be used in this department as it provides a reproducible means of tightly controlling at least one of the most important and variable outcome determinants. Advocates of conventional techniques have stated that other factors such as patient psychological distress, weight, socio-economic status, ethnicity, preoperative function, etc., are far more important than alignment in predicting long-term outcome. Regardless, these factors are fixed and by definition not determined or influenced by the operating surgeon. It can be argued that such statements indirectly emphasize the importance of controlling those determinants of outcome that can be controlled and influenced by the operating surgeon! Thus, whilst in truth the long-term outcome of navigated TKA compared with conventional TKA remains unknown, we adopt a philosophy whereby we prefer to be in control of those outcome determinants that we can reproducibly influence.

Declaration of interest: The authors report no conflicts of interest.