Learning the vagaries of computer-assisted total knee replacement

The Journal of Bone and Joint Surgery. British volumeVol. 87-B, No. 7 AnnotationFree AccessLearning the vagaries of computer-assisted total knee replacementJ. M. Sikorski, M. C. BlytheJ. M. SikorskiConsultant Orthopaedic SurgeonHollywood Specialist Centre, Suite 8/95 Monash Avenue, Nedlands, Western Australia 6009, Australia.Search for more papers by this author, M. C. BlytheOrthopaedic RegistrarRoyal Perth Hospital, Wellington Street, Perth 6000, Western Australia.Search for more papers by this authorPublished Online:1 Jul 2005https://doi.org/10.1302/0301-620X.87B7.15615AboutSectionsPDF/EPUB ToolsDownload CitationsTrack CitationsPermissionsAdd to Favourites ShareShare onFacebookTwitterLinked InRedditEmail In June 2001 the senior author (JMS) was approached by representatives of the Stryker organisation with the request that he introduce and evaluate their computer navigation technology (Stryker Corp, Leibinger, Kalamazoo, Michigan), version 1.1, for total knee replacement. After a visit to Professor Martin Sparmann’s unit in Berlin, and extensive discussions with a representative of the design team, the first computer-assisted total knee replacement (CATKR) of this series was carried out in September 2001. Since then, computer assistance has been used exclusively in the senior author’s practice, with the exception of two cases in which it could not be used for anatomical reasons.The initial request from the Stryker organisation was to introduce the technology into Australia and to evaluate and modify the process. The representatives of the company were aware that the process of evaluation would need to be scientifically rigorous and thus potentially expensive. They were willing to support the research that was required, financially and organisationally. Between June 2001 and September 2001 a programme of clinical research was formulated which has since resulted in a series of publications.1–4The clinical introduction of CATKR required familiarisation with the software and considerable modifications in surgical technique. The greatest intellectual challenge was to design studies which would test the claims of the system. It became rapidly evident that what was required was an objective, quantitative, means of assessing the position of the arthroplasty in all the degrees of freedom that the system was able to control. It was also necessary to establish a comprehensive relational database which would store all the relevant data affecting the short-term alignment, functional outcome and longevity of the implant. The first requirement was met by development of the Perth CT protocol2 which, in conjunction with the Perth alignment index (PAI),4 provided an accurate indicator for alignment of the components. The second need took longer to establish and required the support of another company (Smith & Nephew Inc, Memphis, Tennessee), but has resulted in an Internet-based database with multiple-surgeon input.5In June 2003, the senior author (JMS) was approached by Smith & Nephew to assess the BrainLAB Vector Viscon system, using one of their prostheses. This system differed considerably from that of Stryker and has provided a second learning experience resulting in major modifications of the surgical technique, software and hardware. This article describes the major lessons which have been learned in the process of using these two systems in a total of 192 patients having 232 primary total knee replacements (TKR). Some of the errors and problems which were encountered initially are not discussed because they have been dealt with and are unlikely to recur. However, there are still major challenges and pitfalls in the undertaking of these procedures.Patients and MethodsThe first study using the Perth CT protocol evaluated the Stryker system in cadavers2 and showed that the technology was ready for clinical evaluation. Two parallel clinical series were then initiated. One, a randomised, controlled study which was undertaken by a single surgeon in the final stages of training, has been published.3 The second was a series (CA1) of 122 knees without a control group using version 1.1 of the Stryker software and the Duracon monogram prosthesis (Stryker Corp). There was then a small study (CA2) of seven knees operated on using version 2.0 of the Stryker software. The latter did not require an iliac-crest beacon, and included additional operative options. A third group (CA3) consisted of 103 knees operated on using Brain-LAB software, which went through five versions (β1, β 2, β 3, β 4 and 1.5), and used the Genesis II prosthesis (Smith & Nephew Inc.). The senior author was the sole surgeon in cohorts CA1 to CA3. There was no selection of patients and both primary and revision operations were performed, although only the primary procedures are presented here.Assessment by the Perth CT protocol became routine halfway through the CA1 cohort. Information was gathered on standardised questionnaires and stored in the CAS_WA database5 concerning the details of the patients, the history, physical examination, diagnosis, operative details, progress in hospital after operation and outcome immediately after operation, at six and 12 months and then at yearly intervals.At the same time as cohort CA3 was initiated a prospective series of jig-based TKRs was scanned by CT. This constituted jig-based series JB1. The patients were those of six other surgeons working in the same hospital as the authors and were scanned after operation before discharge. No selection criterion was used other than consent by the patient. Recruitment of patients was discontinued once a surgeon had contributed 30 to 40. A variety of prostheses was used. Finally, a retrospective survey of the patients of the senior author was carried out. They had been operated on in the two years before the start of computer-assisted TKR (1999 to 2000) and had jig-based Genesis II TKRs. They were approached solely because they lived in the Perth metropolitan area. Their clinical status and satisfaction with the procedure varied. They were willing to undergo assessment by CT of their TKR. They constituted series JB2. The characteristics of all the groups are summarised in Table I. The auditing of the CT outcomes of the non-computer-assisted patients was undertaken to justify persistence with what was a new and expensive technology.The overall principles of CATKR have been described.1 They are similar for both systems and are not repeated here.Theoretical and programming considerationsThe fundamental difference between a jig- and a computer-based system is intellectual transparency. It is easy to see how jigs work and to challenge their assumptions. This is not so for computer-based products which are usually used as ‘black boxes’ and this is a major pitfall in CATKR.The systems use infrared sources which produce divergent beams subject to secondary reflections and cameras which possess chromatic aberrations. The position of the objects on which they are focused requires detection software and averaging algorithms. None of these processes was presumed to be perfect. In most cases the resultant accuracy of point detection was either not known or not openly stated.The detection of the centre of rotation of the femoral head was fundamental to defining the mechanical axis. This depended on the presence of a concentric centre of rotation with a relatively stable pelvis. An accuracy read-out was produced, but the exact implication of having an accuracy varying between 0 mm and 4 mm was not clear.The process of registration defined the individual anatomical characteristics of the patient. The surgeon needed to be familiar with the anatomy which was being defined, such as the characteristics and variations of the medial epicondyle. Care needed to be taken to a variable degree. In the BrainLAB system some of the steps involved gathering data for the cosmetic appearance of the images and these were not involved in the alignment of the prosthesis, while others were critical for the positioning of the implant.All software packages share assumptions, use reference axes and reference planes, and have defaults. Those discussed here were based on the assumption that a TKR should be accurately aligned at right angles to the axis of mechanical neutrality. Surgeons who believe that the primary philosophy of a TKR should be soft-tissue balance6 or optimal kinematics should not use this generation of CATKR. While the packages did contain various displays of soft-tissue balance they were not structured to act as an adequate basis for surgery. In the coronal plane, the axis of mechanical neutrality was also the mechanical axis of the limb, which was a line connecting the centre of the femoral head to the centre of the talus. In the sagittal plane a compromise situation was reached because the replaced knee did not always achieve full extension. The axis of mechanical neutrality was taken separately as the mechanical axis of the femur and the anatomical axis of the tibia.There was considerable divergence of opinion as to which reference axes should be used in the axial plane, the rotational axis. For femoral alignment the user was offered a number of choices; the transepicondylar axis, the anteroposterior axis, the posterior condylar axis or an arbitrary mean of the transepicondylar axis and anteroposterior axis. However, the transepicondylar axis is probably the most reliable.7,8 If the wrong choice was made the accuracy of the system was downgraded. For the tibia the problem was greater since no tibial landmarks seemed to have the consistency to provide alignment of the required accuracy. Hence, there was a tendency to prefer making the rotational alignment of the tibial component match the femoral component irrespective of tibial landmarks.9The anteroposterior positioning of the femoral component could be determined mechanically (Stryker Corp.) or by software (BrainLAB). The latter offered a choice between using the anterior cortex of the femoral shaft or the posterior surface of the condyles as a point of reference. Choosing the posterior option involved a considerable risk of overstuffing of the patellofemoral joint or of notching the femur. Anterior referencing was the safer option. It allowed the surgeon to downsize easily when a mistake was made in sizing the component.The depth of the cut or the amount of bone to be removed from the femur and the tibia could be assessed by referencing to the surface of either minimal or maximal wear, the high or low points. Neither was totally satisfactory. Use of the point of minimal wear allowed the cut to be judged to approximate the thickness of the implant if the surface had no wear. If there was wear over the entire bony surface then a compromise was needed. In practice, it was rarely possible both to minimise the amount resected and to remove sufficient bone with a single cut. Our preference was to use the high reference and to take further cuts of 2 to 4 mm as necessary. For the femur it could be argued that the reference should be to the transepicondylar line10 rather than to articular surfaces which are variably worn.Defaults are the manufacturers’ settings, which can be obligatory or optional. There is a tendency for the more complex systems to produce an increasing range of default options. It is necessary to be aware which default settings are present, how to change them and how to check that they have not been changed by accident. At times we encountered defaults which were bizarre and have been difficult or impossible to change.For navigating on targeted or open co-ordinates, the Stryker system provided the surgeon with the absolute coordinates. Thus in aligning the distal femoral cut the display showed a simple stylised outline of the femur and the position of the beacon mounted on the cutting block. The co-ordinates of the beacon were shown as the depth of cut, femoral varus/valgus and femoral flexion/extension. The BrainLAB system provided the same option. It also gave the opportunity to set a plan after which the display showed numbers which related to the plan and not to the initial reference axes. If a complex or erroneous plan was entered then the process of navigation could be very confusing. We have refused to use the targeted approach.There was no calibration process and no information was provided on the impact of variations of the physical organisation of the equipment, such as the distance of the camera from the patient, or the accuracy of registration on the outcome. We did not know if there was any drift in the system. Operative data were gathered by both systems. Version 1.1 of the Stryker system had a reasonable paper print-out which defined the pre-operative deformity, intra-operative bone cuts and the final outcome. In version 2.0 this was slightly downgraded. The BrainLAB system was very poor in this regard until version 1.5. This provided some information, the usefulness of which has not yet been evaluated.Major deformity presented a number of theoretical challenges. The optimal alignment of a TKR is when the prosthesis is aligned along an axis of mechanical neutrality. This goes through the centre of the tibial plateau in the coronal plane and just anterior to the centre in the sagittal plane. However, gross coronal deformity of either femoral or tibial shafts made this impossible. The prosthesis could lie parallel to the axis of mechanical neutrality with it falling to either the medial or lateral side of the centre of the prosthesis (Fig. 1). This presumably produces an unequal distribution of pressure in the polyethylene which could be as damaging as any other form of malalignment.Rotatory deformity, as may occur after a fracture of the femoral shaft (Fig. 2a) or a femoral osteotomy (Fig. 2b), may produce a similar problem. The transepicondylar line may then not be an appropriate guide to the plane of tracking of the patella. Perhaps in such cases soft-tissue balance becomes a better guide to alignment than bony landmarks.Clinical introductionConcentric rotation of the hip and pelvic stability were important in the registration. This technique was impossible in patients who had undergone arthrodesis of the hip and was downgraded and tedious in those with subluxing hips. In the current versions which do not have a pin in the iliac crest, it helped to clamp the pelvis rigidly before draping.The beacon or reflector arrays could be mounted on a variety of anchoring devices, the choice being greater with the BrainLAB system. The Stryker system has a single screw, with a toothed stabilising ring which bites into bone (Fig. 3a). It had to be used in the diaphysis or proximal metaphysis. The teeth produced an ‘apple-corer’ effect and one fracture of the shaft of the femur was associated with its use. BrainLAB hardware had a single screw which could be used in the metaphyseal flare (Fig. 3b) and did not produce the same stress-riser effect. This positioning was very useful in revision surgery since a stem could be introduced past the array. All single-screw mounts were liable to displacement in soft bone. There were also two pin-mounting arrays which were stable even in osteoporotic bone and used wires of small diameter (Fig. 3c). This was preferred in elderly patients but these arrays had to be positioned so that skin and muscle did not tension the wires, thus changing the position of the arrays.The registration process in both systems was software-guided, easy and took less than ten minutes after the first few cases. The Stryker system used battery-powered beacons, which were compact, while the BrainLAB system used reflectors on a mount which was easy to use but mechanically complex and employed reflecting balls which were expensive disposable items. They were more bulky, but could be removed when not being used.Computer assistance could be used either as an adjunct to jigs or without any devices for mechanical alignment. While the former could provide the novice user with a feeling of security it was an unsatisfactory hybrid and was abandoned after ten operations. The ‘freehand technique’ was found to be simpler, quicker and reduced the range of instruments needed for the operation. Initially, the cutting blocks were fixed by two drill bits. These were found to be unstable and we finally resorted to using four smooth pins, two parallel initially and two drilled obliquely. The cutting blocks have been simplified because they did not need jig attachments.The accuracy of the system was totally dependent on the stability of the beacon or reflector arrays. Any movement downgraded the results. Small movements were difficult to detect and once movement of the arrays was recognised clinically the TKR was likely to be grossly malaligned. In both systems there are processes for detecting displacement of the array, but their sensitivity was uncertain and their use was not a mandatory part of the procedure.Intra-operative checks.Once registration and the preparatory surgery were complete the procedure was relatively simple. The jig was set and the cut made and checked. However, aligning the cutting blocks perfectly according to the established landmarks did not inevitably produce perfect bone cuts. It was possible to compare the angle at which the cutting block was set with the angle of the cut which was produced for the three femoral (Fig. 4) and two tibial parameters. Using the available data on all these provided 602 measurements which showed that 508 (84%) of the two angles were within 1° of each other and 590 (98%) within 2°. Two gross individual variations occurred, one of 9° of femoral flexion and another of 6° of femoral rotation. The first of these was associated with loosening of the femoral anchoring screw during surgery. Overall, there was a statistical correlation between the position of the jig and the cuts produced for femoral valgus, flexion and tibial valgus (Spearman correlation coefficients rho = 0.48, 0.62 and 0.54, respectively). No such correlation was found for femoral rotation and tibial slope (rho = 0.24 and 0.30, respectively).Positioning the components on to the cut bone was an additional potential source of error, especially when the bone was very soft. Using uncemented from the could an could have the same effect. positioning a tibial plateau it was possible to control the position and major However, this was not possible with the femoral component since the surface made it impossible to assess alignment without additional time taken for computer-assisted surgery both with experience and with the of by the The iliac-crest made but this has since been and the in the and CA3 cohorts was of the fixed arrays and registration took ten minutes in both systems. was by not having to There was a learning associated with the cutting It was necessary to three changing co-ordinates on a computer while a cutting block and pin but those in surgery should rapidly this was on variations in bone cuts. errors were they could be using jigs errors were not and thus no time was on their in new software has presented major and challenges. The of the iliac-crest pin and from Stryker version 1.1 to version 2.0 made the registration more less accurate and resulted in an in the need for a number of lateral The CT assessment of a very small number that there was a tendency to the femoral cut BrainLAB versions to had default settings which were in version the Stryker series it was possible to compare the final bone cuts as during surgery and the final of the components as shown by CT (Fig. There was some Using the available data on all the provided of the two angles were within 1° of each other and were within 2°. There was no statistical correlation between the cuts produced and the CT measurements for all CT of the CA1 and series showed that the mean was for the for the BrainLAB and for the jig-based system The of perfect = was and the six were at became In all groups the coronal measurements showed the alignment both in of the of a perfect outcome and the range or of results. In the sagittal plane the had a perfect outcome for both components in of cases. In the and series the femoral sagittal alignment was of the same of accuracy. However, tibial sagittal alignment, the tibial was downgraded to of perfect. The range of in tibial was using with and using The in tibial slope were The axial showed between with a to the BrainLAB system over the The of was still with all systems. was not controlled by the The better outcomes of the Stryker system in this their the CT assessment of the and series was in regard to the individual surgeons it was that there was a in outcome. The between surgeons was greater than the mean between the jig-based and computer-assisted The implication of this is that in surgical technique probably had as to as using computer also was that the of the retrospective series which that using computer assistance may have this of and in bone is inevitably associated with some the iliac was being used in the CA1 series there were six to the lateral of the with some three and two both of which were in osteoporotic patients and major was through the of the screw for the and the other through a medial femoral from the of the In the CA3 patients the only have been four in the one of which was very and may have to of the a we the in patients skin is of poor major of computer-assisted TKR is that it has to the alignment more than we have in the The use of CT and the gathering of standardised data that we were able to outcome to alignment in a more use of the Perth CT protocol and the has provided a for which should a to surgical irrespective of we on to the routine use of computer assistance is an technology which has the alignment of It has potential and its to soft-tissue balance and replacement experience the time and additional of the process are reduced to the point at which it is difficult especially there is for mechanical jigs since they are unlikely to achieve more than they The of a jig-based TKR at in by a of of the by the computer technology is to the provided during surgery. However, although it has potential pitfalls in its It is It is a to the and of and company who do not the which they Surgeons need to be aware of the assumptions, the default and the settings which are in before the operation. They need to be very software and when it to over The that it not mean it navigation is a It is potentially and we using I. of the patients in the various of of see for of total knee of the outcome as using the CT protocol and Perth alignment index for all number of with with see for of and and of the measurements for the six in the Stryker and Brain-LAB (CA3) and The outcome is for measurements for the tibial sagittal which according to the design of implant. The into the variable alignment of the tibial slope as do the see for of and and and and and Perth alignment index outcomes of the total knee replacement series by surgeon The series is retrospective while the other series are number of of of a prosthesis in The tibial and femoral components are both within of and yet the mechanical axis of the is to the medial of the This is a of femoral and tibial a deformity resulting from a fracture of the femoral shaft and from a femoral of a Stryker with stabilising teeth which produce an ‘apple-corer’ effect in soft a BrainLAB screw in the of a femur and a BrainLAB in the tibial 4 The between the angle at which the cutting block was set and the resulting bone cut. The minimal was in all cases. The in each block is the The second from the is the mean The third is the and the is the femoral femoral femoral tibial tibial posterior The between the angle of the bone as shown by the computer-assisted system and the resulting alignment of the prosthesis as shown by the CT The in each block is the The is the mean The is the of The in each block is the femoral femoral femoral tibial tibial posterior authors would to the of from the Stryker in the of this and for part in the the of the study M. had with Stryker Australia and Smith & author or one or more of the authors have or for or