Twenty-one patients who had a total of thirty-one consecutive total knee arthroplasties were included in a prospective, randomized, double-blind study at Leiden University Medical Centre. Fixation with cement (the control group), without cement, or with a hydroxyapatite coating (without cement) was randomly chosen during the operation by means of a randomization number. The study was approved by our institutional ethics committee, and the patients gave informed consent.
The components that were fixed with cement had a smooth diamond surface (triangular facets along the surface) on the side that was within bone, whereas the components that were fixed without cement had a mesh-wire surface with or without a hydroxyapatite coating. The window for osseous growth into the mesh-wire surface was 2.25 square millimeters, which corresponds to an equivalent circular pore diameter of 1690 micrometers. All of the prostheses were posterior cruciate ligament-retaining total condylar implants (Interax; Howmedica, Rutherford, New Jersey). This prosthesis and its cast mesh were made of Vitallium (cobalt-chromium) alloy. The hydroxyapatite coating was vacuum plasma-sprayed onto the component after shot-blasting. The hydroxyapatite was at least 90 percent pure, and it had a crystallinity of more than 90 percent, a mean density of 90 ± 2 percent, and a mean thickness of 60 ± 30 micrometers.
A patellar component made of ultra-high molecular weight polyethylene was used in all but one total knee arthroplasty. No patellar component was used in one knee because the remaining patellar bone was too thin. The operation was performed through a standard mid-line incision with a medial parapatellar arthrotomy. After soft-tissue releases, the prosthesis was implanted according to the manufacturer's instructions. When bone cement was used, the cut bone surfaces were mechanically pulse-lavaged with a device manufactured by Zimmer (Warsaw, Indiana) before Palacos bone cement (Schering, Kenilworth, New Jersey) was applied.
Postoperatively, all of the knees were fitted with a Jones bandage. Weight-bearing and flexion exercises were avoided until the first roentgenogram had been made for roentgen stereophotogrammetric analysis (between one and five days postoperatively).
The patients were evaluated preoperatively; at one week, three weeks, six weeks, three months, six months, and one year postoperatively; and at yearly intervals thereafter. At each evaluation, the clinical (Knee Society9) score was assessed and roentgenograms for roentgen stereophotogrammetric analysis were made. Immediately after the operation and at the two-year follow-up evaluation, standard anteroposterior and lateral roentgenograms of the knee as well as axial roentgenograms of the patella were made with the patient standing, as described by the Knee Society5.
There were no significant differences among the three groups with regard to age (mean [and standard deviation] , 68 ± 11.6 years; p = 0.1, Kruskal-Wallis test), body-mass index (mean, 23 ± 2.8 kilograms per square meter; p = 0.8 Kruskal-Wallis test), or stage of osteoarthrosis (mean, 4 ± 2.4 according to the classification system of Ahlbäck1 and 5 ± 0.6 according to that of Larsen et al.11; p = 0.7, Kruskal-Wallis test). Osteoarthrosis was diagnosed in five knees and rheumatoid arthritis, in twenty-six knees. One patient (one knee) died after the one-year evaluation, before the final follow-up evaluation for the study. One knee in a patient who had rheumatoid arthritis and severe osteoporotic bone intraoperatively could not be used for roentgen stereo-photogrammetric analysis because several markers had migrated. Thus, thirty knee prostheses were evaluated with roentgen stereophotogrammetric analysis and thirty-one knee prostheses, with conventional roentgenograms. Eleven total knee components fixed with cement, ten hydroxyapatite-coated components fixed without cement, and ten noncoated components fixed without cement were studied.
During the knee arthroplasty, six, seven, or eight bone-markers (one-millimeter-diameter tantalum balls [Industrial Techtonics, Ann Arbor, Michigan]) were inserted into the tibial metaphysis. Three two-millimeter-diameter Vitallium markers had been attached to the tibial component by the manufacturer (Howmedica).
The setup for the roentgen stereophotogrammetric analysis consisted of two synchronized roentgen tubes positioned approximately 1.5 meters above the roentgen film, with each tube projected at one-half of the film (total area of the film, thirty-five by forty-two centimeters). The exposure of the roentgen film by both roentgen tubes was simultaneous. The angle of each roentgen tube with the vertical was 20 degrees. Furthermore, a Plexiglas calibration box with twenty-six one-millimeter-diameter tantalum markers on the top and bottom layers was used. The positions of the markers were determined precisely by a mechanical measuring device that had an accuracy of 0.001 millimeter. These box-markers define the coordinate system and determine the roentgen foci positions. The patient is positioned above the calibration box. The coordinates of the bone-markers and prosthesis-markers on the roentgenogram (Fig. 1) were recorded on a measurement table that had a video camera (Philips, Eindhoven, The Netherlands) with a forty-power-magnification microscope (Wild GMBH, Heerbrugg, Switzerland) attached. The coordinates were determined with two digital linear gauges (x and y axes) that were part of the measurement table, and they were then processed with roentgen stereophotogrammetric analysis software that was developed at the Department of Orthopaedic Surgery at our institution. Basically, this software determined the three-dimensional position of the prosthesis-markers in relation to the bone-markers with rigid-body kinematics. Thus, micromotion throughout the follow-up period was measured with a high degree of accuracy. Unstable markers were detected and eliminated from the analysis. The stability of individual markers within each segment was assessed by minimization of the mean error of rigid-body fitting. (Ideally, the difference in distances between the tantalum bone-markers should be the same at successive follow-up moments.) Unacceptably high values of differences between these distances (that is, a mean difference of more than 0.3 millimeter compared with the distance on the first roentgenogram) indicates instability of those markers; consequently, they were omitted from further calculations. During this study, a total of five markers were labeled as unstable. Three of the unstable markers were in a patient who had severe osteoporosis, as mentioned.
The first roentgen stereophotogrammetric analysis served as the reference baseline. All subsequent evaluations of micromotion were based on the relative position of the prosthesis with respect to the bone at that time. Migration of the tibial component was expressed as translatory movements along the three orthogonal axes: longitudinal (y axis), transverse (x axis), and sagittal (z axis). The reproducibility of the roentgen stereophotogrammetric analysis was determined with repeat measurements on thirty roentgenograms. The standard deviations thus determined were 0.03 millimeter for translations along the x axis, 0.08 millimeter for translations along the y axis, and 0.13 millimeter for translations along the z axis.
In contrast to other studies8,19, in the present study micromotion was recorded along the three orthogonal axes instead of as maximum total point motion because the direction of motion is unclear when only the maximum total point motion is given. Because a prognostic value can be attributed to a maximum total point motion of 0.2 millimeter or more during the second post-operative year19, values for micromotion of individual components along each of the three orthogonal axes were subdivided into two groups: those indicating stability (less than 0.2 millimeter of micromotion) and those indicating instability (0.2 millimeter of micromotion or more).
Statistical Analysis
Mean values and standard deviations were calculated for all variables. For a comparison of the mean values in the three groups, a Kruskal-Wallis test was used. A repeated-measures one-way analysis of variance was performed on the micromotion data from the roentgen stereophotogrammetric analysis to determine differences between the three types of prosthetic fixation throughout the two-year follow-up period.
Clinical Results
Total knee arthroplasty took longer with cementing of the tibial component (134 ± 10.2 minutes) than it did with fixation of either the hydroxyapatite-coated component (120 ± 9.0 minutes) or the noncoated component without cement (120 ± 13.8 minutes) (p < 0.05, Kruskal-Wallis test).
The preoperative functional scores (mean, 10 ± 2.9 points) and knee scores (mean, 24 ± 3.2 points), according to the system of the Knee Society, did not differ significantly among the three groups (p > 0.1, Kruskal-Wallis test). At the two-year follow-up evaluation, these values were 41 ± 8.3 and 79 ± 3.2 points, respectively; again, no significant difference could be detected among the three groups, with the numbers available (p > 0.1, Kruskal-Wallis test). The mean preoperative flexion contracture for the three groups was 27 ± 4.9 degrees, which had decreased to 8 ± 3.3 degrees at the two-year follow-up evaluation. The mean subscore for mediolateral instability, according to the system of the Knee Society, decreased from a category of 1.8 ± 0.2 preoperatively to one of 1.2 ± 0.02 at the time of the two-year follow-up evaluation. (Category one indicates less than 5 degrees of mediolateral instability and category two, 6 to 9 degrees of instability.)
Routine roentgenograms of the knee revealed no radiolucent lines of two millimeters or greater in thickness along the tibial, femoral, or patellar component in any of the three groups at the two-year follow-up evaluation.
The mean preoperative femorotibial angle for the three groups was 173 ± 1.9 degrees (a valgus deformity). At the two-year follow-up evaluation, the mean angle was 176 ± 0.7 degrees, and no significant difference (p = 0.6) was detected among the three groups, with the numbers available.
Complications
One infection (with Staphylococus aureus, secondary to a prepatellar bursitis) developed six months postoperatively in a knee with a noncoated component fixed without cement. Operative intervention included removal of the prosthesis and débridement, and the patient was managed with six weeks of antibiotic therapy. An arthrodesis with two plates was subsequently performed because of a rupture of the patellar ligament.
Roentgen Stereophotogrammetric Analysis of Translation of the Components
Micromotion, expressed as movement along the three orthogonal axes, varied considerably among the individual components. The largest amount of micromotion was seen along the longitudinal axis: a noncoated tibial component fixed without cement had subsided 2.3 millimeters at the two-year follow-up evaluation. This particular noncoated component also had the largest values for micromotion along the sagittal and transverse axes: 1.0 and 1.8 millimeters, respectively.
Longitudinal Axis (Fig. 2 and Table I)
At three weeks postoperatively, we were not able to detect a significant difference in mean subsidence among the three groups (p = 0.8, Kruskal-Wallis test). At the two-year follow-up evaluation, the components fixed with cement had subsided a mean of -0.05 ± 0.109 millimeter along the longitudinal axis; the noncoated components fixed without cement, a mean of -0.73 ± 0.924 millimeter; and the hydroxyapatite-coated components fixed without cement, a mean of -0.06 ± 0.169 millimeter.
Because differences in trends, with regard to micromotion, among the three groups are more important than differences in absolute mean values at the follow-up evaluation13, the values for micromotion were compared throughout the two-year period. During the first two postoperative years, no significant difference was found between the mean subsidence (translation along the longitudinal axis) of the hydroxyapatite-coated components and that of the cemented components (p = 0.9, analysis of variance). However, the noncoated components subsided significantly more (p < 0.05, repeated-measures analysis of variance) than did both the hydroxyapatite-coated components and the cemented components.
During the second postoperative year, four noncoated components subsided at least 0.2 millimeter along the longitudinal axis.
Sagittal Axis (Fig. 3 and Table I)
We were not able to detect a significant difference in the values for micromotion along the sagittal axis among the three groups at three weeks postoperatively (p > 0.05, Kruskal-Wallis test). At the two-year follow-up evaluation, the mean micromotion of the noncoated, hydroxyapatite-coated, and cemented components was 0.65 ± 0.662, 0.05 ± 0.09, and 0.03 ± 0.190 millimeter, respectively.
We detected no significant difference (p = 0.5, repeated-measures analysis of variance) between the hydroxyapatite-coated components and the cemented components with regard to the trends of micromotion along the sagittal axis during the first two postoperative years. However, the difference between the noncoated components and the cemented components was significant (p = 0.01, repeated-measures analysis of variance) and that between the noncoated components and the hydroxyapatite-coated components was also significant (p = 0.03, repeated-measures analysis of variance) during the two-year follow-up evaluation.
During the second postoperative year, five non-coated components, three cemented components, and one hydroxyapatite-coated component migrated at least 0.2 millimeter along the sagittal axis.
Transverse Axis (Fig. 4 and Table I)
During the two-year follow-up evaluation, no significant difference (p = 0.9, analysis of variance) was found with regard to the trend of micromotion along the transverse axis between the tibial components fixed with cement and the hydroxyapatite-coated tibial components fixed without cement. The mean values for micromotion at successive follow-up evaluations were significantly lower (p < 0.01, repeated-measures analysis of variance) for the cemented components and the hydroxyapatite-coated components than they were for noncoated components.
During the second postoperative year, only one noncoated component migrated at least 0.2 millimeter along the transverse axis.
Because of its accuracy10,19,20, roentgen stereophotogrammetric analysis is a valuable tool for the evaluation of new prosthetic implants and fixation techniques with regard to their actual performance (stability of fixation) in vivo. The validity of roentgen stereophotogrammetric analysis has been stressed recently because extensive micromotion at two years was associated with long-term prosthetic loosening10,19. In two studies10,19, all of the components that had been revised at the ten-year follow-up evaluation had had progressive, continuous micromotion. However, merely measuring micromotion without any specific scientific purpose is senseless.
As we expected, on the basis of other studies on knee prostheses9,12,13,16,26,27, our patients improved clinically after total knee arthroplasty. However, results as measured by the Knee Society score (or by any clinical score for that matter) vary somewhat among studies because (among other reasons) the score reflects not only the involved knee but other patient-related factors as well. Most of the patients in our study had rheumatoid arthritis with involvement of multiple joints, so the clinical results reflected not only the involved joint but also the total disability. Thus, patients who have rheumatoid arthritis rarely, if ever, regain the maximum possible clinical knee score postoperatively. With the numbers available, the Knee Society scores did not differ significantly among the three groups, but the discriminative power of those scores is too low for differentiation between fixation techniques. Thus, the amount of micromotion was the actual point of interest in the present study.
Loosening of total knee prostheses seems to start with the tibial component; therefore, research addressing micromotion of tibial components fixed in different ways is of interest. In the present study, all of the tibial components migrated, irrespective of their mode of fixation. However, it is important to realize that initial postoperative migration (movement over time) is not necessarily equivalent to poor prosthetic fixation because all prostheses probably have a postoperative micromotion phase. After prosthetic implantation, bone resorption initially occurs as a result of the operative trauma (slight bone necrosis from cementing or from sawing and reaming). However, in some patients this initial micromotion continues and is progressive after the first or second postoperative year; thus, there is no prosthesis-settling phase. In one study, prostheses that were loose at the ten-year follow-up evaluation were found to have had a higher prevalence of progressive micromotion, as well as a greater magnitude of micromotion, in the second postoperative year than prostheses that were well fixed at ten years19. In the present study, at least 0.2 millimeter of micromotion, which continued after the first follow-up period, occurred mainly in the group of noncoated components fixed without cement, and, because of their higher likelihood of loosening compared with components that have less micromotion19, these components should be followed closely in the future. Other investigators have also reported a higher rate of micromotion of noncoated prostheses fixed without cement19.
It has been postulated that an increased rate of loosening occurs in knees that are tight in extension (a residual flexion contracture) or as a result of instability of the knee joint (the so-called teeter-totter phenomenon)25. There was no notable flexion contracture or instability of any knee in the present study. Because the clinical scores and subscores (for range of motion, stability, and so on) were not significantly different (with the numbers available) among the three study groups, the fixation technique was the major determinant of the stability and instability of the components, as well as the continuous migration of the components, at the two-year follow-up examination. In an experimental study of fourteen dogs, a fibrous membrane developed around all of the titanium-coated and hydroxyapatite-coated titanium implants after four weeks of immobilization22. Continuous micromotion after this period resulted in development of bone around the hydroxyapatite-coated implants, whereas fibrous connective tissue remained around the titanium-coated implants. The so-called stabilized migration (no continuous, progressive migration of 0.2 millimeter or more) of most of the hydroxyapatite-coated and cemented implants in the present study suggests a strong implant-prosthesis interface (osseous ingrowth or good cement-bone interdigitation).
In the present study, which largely comprised patients who had rheumatoid arthritis, noncoated tibial components fixed without cement had significantly more micromotion than did noncoated tibial components fixed with cement and hydroxyapatite-coated tibial components fixed without cement. The rather large window for osseous growth into the surface of the components fixed without cement (2.25 square millimeters) might not encourage osseous ingrowth. The addition of hydroxyapatite not only decreases the size of the gaps because it is plasma-sprayed onto the surface, it also enhances osseous growth into gaps as wide as two millimeters21. Furthermore, it can transform motion-induced fibrous tissue into bone22. A recent study of dogs15 showed that hydroxyapatite inhibits the migration of polyethylene particles along titanium implants because it seals the bone-prosthesis interface. This might have a positive effect on the long-term survival of the prosthesis. One concern about the hydroxyapatite coating is, its degradability because the production of particles might cause third-body wear and loss of mechanical fixation2. However, this theoretical disadvantage has been poorly documented. Furthermore, because hydroxyapatite has osteoconductive properties, bone is present at the prosthetic interface when the hydroxyapatite dissolves.
In conclusion, tibial components fixed with cement and hydroxyapatite-coated tibial components fixed without cement have far less micromotion along the three orthogonal axes than do noncoated tibial components fixed without cement. On the basis of the results of this study, which was largely composed of patients who had rheumatoid arthritis, it appears that the addition of the biological mediator hydroxyapatite provides adequate fixation to prevent mechanical loosening of the tibial component of a total knee arthroplasty.