The importance of optimal alignment in total knee arthroplasty is related to evidence that poor alignment is associated with early loosening1,2, patellofemoral disorders3,4 and uneven wear of the polyethylene liner1,5, which result in limitation of function6 and ultimately failure of the prosthesis1,2,7.
Currently, three main surgical techniques are used to facilitate correct alignment of components in total knee arthroplasty: intramedullary guides for both the femur and the tibia, an intramedullary guide for the femur and an extramedullary guide for the tibia, and computer-assisted surgery for both the femur and the tibia.
There is a paucity of Level-I evidence on the comparative benefits of intramedullary and extramedullary tibial guides. The few randomized trials conducted have had conflicting results, and to our knowledge none have involved the use of computed tomography (CT) analysis. One trial demonstrated improved alignment with intramedullary guides8; another, with extramedullary guides9; and a third, little difference between the two types of guides10. Authors of observational studies have compared intramedullary guides with extramedullary guides, with some concluding that the intramedullary technique is more accurate for determining correct alignment11,12 and others reporting little difference between the two techniques13-16. Orthopaedic surgeons tend to use the technique with which they are most comfortable rather than the technique that is most accurate or for which there is the best supporting evidence17.
The advent of computer-assisted surgery has facilitated more accurate component positioning. A meta-analysis of thirty-three studies, including eleven randomized controlled trials, provided evidence that computer-assisted surgery produces a tighter spread of deviations from neutral alignment of the coronal mechanical axis than do conventional surgical techniques with intramedullary and extramedullary guides18. However, computer-assisted surgery was also associated with increased operative time and cost. The authors concluded that there was no evidence of an overall benefit of computer-assisted surgery, as the methodological weaknesses of the available trials precluded reliable conclusions. Subsequent randomized controlled trials have either shown improved alignment with computer-assisted surgery as compared with conventional surgery19-21 or no difference between the two22,23. Questions about which surgical technique is the best remain.
The goal of our study was to determine the most accurate technique for component alignment in total knee arthroplasty by comparing computer-assisted surgery with intramedullary and extramedullary techniques.
Ethics approval was granted by the Hospital Human Research and Ethics Committee. The trial was registered with the Australian New Zealand Clinical Trials Registry (ACTRN12609000404224). The study design and reporting were based on the CONSORT (Consolidated Standards of Reporting Trials) principles.
Eligibility Criteria
All patients admitted under the care of one surgeon for a primary total knee replacement to treat knee osteoarthritis and varus deformity were considered for inclusion in the trial. Patients were excluded from the trial if they had valgus deformity of the knee or tibial deformity from a previous fracture and/or osteotomy.
Setting and Location
The trial was conducted at two metropolitan tertiary-care hospitals.
Interventions
Patients underwent total knee arthroplasty with use of one of three methods for tibial alignment: an extramedullary guide (the extramedullary guide group), an intramedullary guide (the intramedullary guide group), or computer-assisted surgery (the computer-assisted surgery group). The Genesis-II Total Knee System (Smith & Nephew, Memphis, Tennessee) was used, in accordance with the manufacturer's instructions, for both the arthroplasties done with the intramedullary tibial guide and the ones done with the extramedullary tibial guide. Femoral alignment was performed with an intramedullary guide in both procedures (i.e., those done with the intramedullary tibial guide and those done with the extramedullary tibial guide). The computer-assisted surgery technique was used, for both femoral and tibial alignment, with utilization of BrainLab Knee Essential software (BrainLab, Feldkirchen, Germany) for the Genesis-II system.
All surgery was performed by one experienced knee arthroplasty surgeon (R.J.K.K.) who was familiar with all techniques. The procedures were standardized by the use of the same medial parapatellar surgical approach, a cemented Genesis-II prosthesis, and an intraoperative pneumatic tourniquet in all groups.
Outcomes
Anteroposterior hip-to-ankle standing radiographs were made six weeks postoperatively. For these radiographs, the feet were placed apart with the knees in maximum extension and the toes pointing straight. The patella was placed in the direction of the x-ray source as a rotation guide. The primary outcome was the coronal tibiofemoral angle, defined as the angle between the femoral and tibial mechanical axes of the lower extremity (Fig. 1). Deviations from the neutral angle of 180° were recorded as positive values if varus was present and as negative values if valgus was noted. Statistical analysis of measurements of malalignment was performed with use of absolute values. Up to 3° of deviation from neutral alignment was considered acceptable, whereas values outside of this range were classified as outliers. All radiographs were viewed with use of a high-resolution monitor and IMPAX radiograph viewing software (Agfa Healthcare, Mortsel, Belgium), and the angles were calculated with use of a digital vectorial goniometer.
Two independent investigators blinded to the trial interventions performed the radiographic measurements. The mean of the two results was used in the analysis.
At three months after the total knee arthroplasty, CT scans were made with the Perth CT protocol24. All scans were performed at a single radiology center and were interpreted by a single radiologist blinded to the trial intervention. The coronal tibiofemoral angle (the primary outcome) and the anatomical axes of the tibia and femur were identified, on the basis of three-dimensional landmarks, in both the anteroposterior and the lateral plane. The coronal and sagittal alignment of the prosthesis was then measured against these axes. The rotation of the femoral component was measured relative to the transepicondylar axis. The tibiofemoral mismatch angle was calculated by superimposition of the femoral and tibial axial images.
Other recorded outcomes were postoperative complications and the operative time from skin incision to application of the dressing.
Randomization
Patients had an equal probability of being randomly assigned to each of the three arms. Randomization was carried out with use of a computer-generated list with random permuted blocks of three. The randomization sequence was concealed prior to enrollment and was not made available until the morning of the surgery.
Blinding
Both the patients and the investigators measuring the final outcomes were blinded with regard to the trial arm to which the patient had been assigned.
Statistical Methods
A linear mixed model was fitted to the data with a fixed effect of randomization group, and a random effect of each patient within the group. This random effect takes into account the correlation between the different raters’ measurements. Age and sex as well as their interaction were included in all of the models. A logistic regression model was fitted to the data indicating the proportion of alignments that were outside of the allowable range, to determine whether this proportion differed between the groups. A p value was determined with analysis of variance (ANOVA) to test for significant differences in the amounts of malalignment among the randomization groups. The intraclass correlation was also calculated for the randomization groups to assess the consistency of the two investigators with regard to measuring the alignment variables.
Sample Size
The results of a pilot study carried out prior to the definitive study showed a trend for computer-assisted surgery resulting in greater accuracy in the coronal tibiofemoral angle. Assuming a difference of 1.8° in the coronal tibiofemoral angle between groups with a standard deviation of 2.3°, we performed a power analysis with α = 0.05 (one-sided significance level) and 1 − β = 0.9. We determined that twenty-eight patients per group—a total of eighty-four patients—was needed to achieve significance. Our goal was to recruit a minimum of 100 patients to allow for losses to follow-up.
Source of Funding
Funding for this trial was provided by The Hollywood Private Hospital Research Foundation and Smith & Nephew Surgical Pty Ltd.
A total of 107 patients were randomized for the trial. Thirty-six were randomized to the computer-assisted surgery group; thirty-five, to the extramedullary guide group; and thirty-six, to the intramedullary guide group. Seven patients were unavailable for CT, and one of the seven also did not have radiographic studies. One patient randomized to the computer-assisted surgery group had surgery with extramedullary guides because of failure of the computer-assisted surgery software. The patient flow is summarized in Figure 2. The three groups were similar with regard to age and sex (see Appendix).
The two independent assessors showed good interobserver reliability, with an intraclass correlation of 91.9% for measurement of the coronal tibiofemoral angle. The intraclass correlation between the radiographic and CT measurements of the coronal tibiofemoral angle was only 46.0%.
Radiographic Results
Measurement of the weight-bearing coronal tibiofemoral angle on the anteroposterior hip-to-ankle standing radiographs demonstrated improved accuracy with computer-assisted surgery (mean malalignment, 1.61°) as compared with the extramedullary (3.03°) and intramedullary (2.86°) guide groups. The proportion of patients with a coronal tibiofemoral angle outside the acceptable range of ±3° was compared among the groups, and significantly fewer outliers were found in the computer-assisted surgery group (19%; seven of thirty-six) than in the intramedullary (36%; thirteen of thirty-six) and extramedullary (38%; thirteen of thirty-four) guide groups (Fig. 3).
The range, mean, and standard deviation of each of the CT measurements in the randomized groups are presented in Table I. Only the coronal tibiofemoral angle differed significantly among the three groups (p = 0.007), with significantly less malalignment in the computer-assisted surgery group (mean, 1.91°) as compared with the values in the extramedullary (3.22°) and intramedullary (2.59°) groups. The distribution of the coronal tibiofemoral angles is shown in Figure 4. All other measurements showed a trend toward greater accuracy in the computer-assisted surgery group, although these trends did not reach significance. There was no significant difference in any other measurements between the intramedullary and extramedullary guide groups.
Operative Time
There was a significant difference in the duration of surgery among the three groups (p < 0.0001). The mean operative time in the computer-assisted surgery group (107 minutes) was significantly higher than that in the extramedullary guide group (eighty-three minutes) or the intramedullary guide group (eighty minutes). There was no significant difference in operative time between the extramedullary and intramedullary guide groups.
Complications
There was no significant difference in postoperative complications among the three groups. There was one case of pulmonary embolism and one deep infection requiring irrigation and a change of the liner in the extramedullary guide group. Two patients (one in the extramedullary and one in the intramedullary guide group) required manipulation under anesthesia because of knee stiffness.
This study provides evidence that radiographically measured alignment after computer-assisted surgery is significantly better than that after surgery performed with the use of conventional intramedullary or extramedullary guides. This increase in accuracy with computer-assisted surgery comes at a cost of increased operative time. There was no significant difference in any of the measured outcomes between the intramedullary and extramedullary guide groups.
Numerous investigators have evaluated the long-term outcomes of total knee arthroplasty and have compared radiographic measurements of alignment. A coronal tibiofemoral angle in excess of ±3° has been associated with worse functional outcomes and higher rates of implant failure19,25-28. It is imperative that alignment within this range be achieved for optimal outcomes from total knee arthroplasty.
Previous randomized control trials comparing conventional total knee arthroplasty with computer-assisted surgery have demonstrated a smaller range of deviation from the tibiofemoral alignment in the coronal plane and fewer outliers (in excess of ±3°) with computer-assisted surgery18-21,29,30. Other studies, however, have failed to show a significant difference23,24. One would expect that the more familiar the surgeon is with the technique that he or she is using, the better the outcomes. Although the surgeon who performed all of the operations in the present study was well-versed in all three techniques, he had less experience with computer-assisted surgery; yet that procedure produced significantly better outcomes.
Several observational studies11-17 as well as some randomized controlled trials8-10 have compared intramedullary and extramedullary guides, with conflicting results. There may be several possible explanations for the differences in the findings of these studies. Some included too few participants to detect any difference between techniques or were of poor methodological quality11-18, and many used only radiographs and not CT scans8-18,20,30.
We believe our study to be the first adequately powered randomized clinical trial to compare all three techniques, with CT verification of alignment. The accuracy of measurements on radiographs is affected by rotational positioning of the limb as well as by the presence of any knee flexion contraction, errors that can be eliminated with CT evaluation. Both radiographic and CT measurements provide useful information, but neither alone is sufficient. CT may be more accurate, but standing coronal radiographs demonstrate the limb alignment on loading, which may differ from that without loading. The low intraclass correlation between the radiographic and CT measurements may be due to actual differences between the values measured in the weight-bearing and supine positions of the patient. Undoubtedly, the difficulty of reproducing limb positioning at the time of imaging contributes substantially to inaccuracy.
The percentages of patients in this study who had greater than ±3° of malalignment following the total knee arthroplasties with the conventional methods were surprisingly high (36% and 38%). These disappointing results are similar to those in other studies, in which 22% to 64% of patients had tibiofemoral alignment outside of what is considered acceptable19-21,29,30. In the studies that demonstrated no benefit from computer-assisted surgery, much lower percentages of patients (17.5% to 18%) had malalignment outside of the acceptable range after conventional surgery22,23. Although the authors of these studies reported no significant difference between computer-assisted surgery and conventional surgery, there was a trend toward better alignment with computer-assisted surgery. However, these studies may have been underpowered to detect a difference between techniques.
Despite the increasing evidence that computer-assisted surgery produces better implant alignment18-21,29,30, most surgeons continue to use conventional implant alignment techniques. However, most of the studies comparing computer-assisted total knee arthroplasty with conventional total knee arthroplasty have been published only recently. Surgeons prefer the method with which they are most familiar, and operating with computer-assisted surgery requires learning new skills. Computer-assisted surgery is associated with increased operative time and increased cost, both of which may be additional disincentives to use computer-assisted surgery. This study adds to the body of evidence demonstrating that computer-assisted total knee arthroplasty has better radiographic outcomes than conventional surgery with extramedullary and intramedullary guides. As we are not aware of any studies showing improved radiographic outcomes following total knee arthroplasties done with conventional intraoperative guides, we suggest that computer-assisted total knee arthroplasty should be accepted as providing the most accurate alignment of implants. Long-term clinical studies are required to determine if this improvement in radiographic accuracy translates into improved patient function and prosthetic survival.