68 consecutive posterior cruciate retaining TKAs (Interax; Howmedica, Rutherford, NJ) performed in 48 patients because of osteoarthritis or rheumatoid arthritis, were included in a randomized, controlled trial in an academic hospital between 1993 and 1998. The study was done in compliance with the Helsinki Declaration and was approved by the institutional ethics committee (pp 166/93; November 30, 1993), and patients gave informed consent. 24 TKAs were performed with cemented tibial components, 24 with HA-coated tibial components, and 20 with uncoated tibial components (Table 1).

Reporting was in accordance with the CONSORT guidelines and the RSA guidelines (^{Moher et al. 2001},^{Valstar et al. 2005}). 2-year migration results and details of patients and methods have been reported previously (^{Nelissen et al. 1998}).

Fixation of the tibial component with HA was compared to uncoated fixation and to fixation with cement. The inserts were made of ultra-high-molecular-weight polyethylene (UHMWPE), sterilized by gamma radiation in air, and machined from ram-extruded GUR 415 resin containing calcium stearate.

In the present study, patients were followed for 11–16 years, or until death or revision of the tibial component (Figure 1). To account for the learning curve with this—at the time—new TKA at our institution (1992) and to gain experience with the RSA equipment, the first 12 TKAs were not randomized and received cemented fixation. These 12 TKAs were not included as part of the study. Nevertheless, RSA analysis was performed in order to exclude potential selection bias for the consecutive study. The migration was similar to that of the randomized cemented cases (p = 0.3), as analyzed with a generalized linear mixed model. During the study, patients remained blind regarding the fixation method. Observers were blinded during the RSA analysis regarding the presence or absence of HA coating, so the study was double-blind regarding the type of uncemented fixation (HA-coated or uncoated). Since cement is visible on (RSA) radiographs, the study was single-blind regarding the comparison with cement.

All TKAs were performed by two experienced knee surgeons or under their direct supervision, and implanted through a standard midline incision and medial parapatellar arthrotomy. 6–8 tantalum markers were inserted into the tibial metaphysic bone before final implantation of the tibial component.

In the cemented group, Palacos bone cement (Schering, Kenilworth, NJ) was used after mechanical pulse-lavage of the cut bone surfaces. To allow migration measurements by marker-based RSA, three 2-mm Vitallium markers had been attached to the tibial component by the manufacturer.

The RSA technique has been described previously (^{Nelissen et al. 1998}). Analysis of the RSA examinations was performed with MBRSA 3.2 software (Medis Specials, Leiden, the Netherlands). The marker configuration model RSA technique was used for measurement of the pose of a rigid body in situations where less than 3 markers could be detected in both images of an RSA radiograph (^{Kaptein et al. 2005}). In 2002, the calibration cage was replaced. Accuracy of the RSA set-up prior to 2002, as determined by double examination analysis (n = 40), was as follows for the translations expressed in means: x-axis 0.00 mm (SD 0.07 mm), y-axis 0.01 mm (SD 0.06 mm), and z-axis –0.02 mm (SD 0.13 mm) (^{Nelissen et al. 1998},^{Valstar et al. 2005}). From 2002 onwards (n = 44), the accuracy was: x-axis 0.00 mm (SD 0.03 mm), y-axis 0.01 mm (SD 0.06 mm), and z-axis –0.01 mm (SD 0.08 mm) according to ^{Kaptein et al. (2007)}. These values indicate a high level of precision for the measurement of migration of the tibial component relative to the bone and absence of any systematic bias.

Weight-bearing and flexion exercises were postponed until after the first RSA radiograph (1–5 days postoperatively). The patients were evaluated both clinically and by RSA examinations at predefined follow-up times (3 weeks, 6 weeks, 3 months, 6 months, and 1 year postoperatively) and then on an annual basis.

The first RSA radiograph served as a baseline reference for the migration measurements.

Maximal total point motion (MTPM)—migration of the point on the prosthesis that has moved the most—was used to determine whether the groups were different regarding migration. When MTPM was different between the groups, translations and rotations along the x-, y-, and z-axis were evaluated to determine how they differed.

Most migration occurs in the first postoperative year, followed by either stabilization or continuous migration of the tibial components (^{Ryd et al. 1995}). Since MTPM represents the length of a vector, which cannot be subjected to regular addition or subtraction, an additional RSA analysis was carried out with the 1-year postoperative RSA radiograph as a reference.

Clinical evaluation was performed according to the Knee Society score (KSS) and Hospital for Special Surgery score (HSS) at each follow-up (^{Insall et al. 1989}).

In addition to the RSA radiographs, conventional weight-bearing radiographs were acquired at 2-, 5-, 10-, and 15-year follow-up and graded according to the Knee Society roentgenographic evaluation: femoral-tibial aligment (FTA angle) and also alfa angle (frontal angle of the femoral component), beta angle (frontal angle of the tibial component), and delta angle (sagittal angle of the tibial component) (^{Ewald 1989}).

Due to the high degree of accuracy of RSA, 20 TKAs were required for each trial arm—as was standard for RSA studies at the time the present study was designed (1992), (^{Albrektsson et al. 1992}, ^{Grewal et al. 1992}). The results were analyzed according to the intention-to-treat principle. To take into account the repeated-measures design of the study, any missing follow-up occasion, variation in duration of follow-up, and bilaterality, and also to allow for confounder correction (^{Pijls et al. 2011}), a generalized linear mixed model (GLMM) was used (R software version 2.12.0), which is considered to be the primary analysis method for this type of clinical study (^{DeSouza et al. 2009}). In accordance with recent studies, a log-transformation was used for maximal total point motion (MTPM)—migration of the point on the prosthesis that has moved the most—because it is not normally distributed (^{Astephen Wilson et al. 2010}). Due to multiple primary outcomes (translations, rotations), a Holm-Bonferroni correction for multiple testing was performed (Holm 1979). Means are presented until 10 years of follow-up. Afterwards, cases are presented individually. 95% confidence intervals (CIs) were calculated.

The migration analysis was composed of 742 RSA analyses using the direct postoperative RSA radiograph as reference. Figure 1 shows the number of valid RSA examinations for each follow-up occasion. Figure 2 shows the mean migration expressed in MTPM for each fixation group up to 10 years postoperatively. Throughout the follow-up period, the uncoated tibial components showed mean 0.39 mm (95% CI: 0.16–0.62) more migration than the HA-coated tibial components and mean 1.0 mm (CI: 0.82–1.18) more than the cemented tibial components, while the HA-coated components migrated mean 0.61 mm (CI: 0.42–0.80) more than the cemented components (unadjusted: p < 0.001, GLMM; and adjusted for age, sex, diagnosis, revisions, and BMI: p < 0.001, GLMM). The mean migration at 10 years was 1.61 mm for the osteoarthritis patients and 1.52 mm for the rheumatoid arthritis patients (p = 0.2, GLMM adjusted for fixation, age, sex, revision, and BMI).

To determine whether migration patterns varied between the groups, the mean translations and rotations were determined. The uncoated tibial components showed statistically significantly increased subsidence, external rotation, and lateral and anterior translation (in the order of clinical relevance). The addition of HA affected migration by decreasing subsidence by mean 0.26 mm (CI: 0.10–0.42) and external rotation by mean 0.47 degrees (CI: 0.27–0.67) compared to uncoated components.

463 RSA analyses composed the migration analysis relative to the first postoperative year (Figure 3). There was a statistically significant difference between the fixation groups regarding migration from 1 to 10 years (unadjusted: p < 0.001, GLMM; and p < 0.001, GLMM adjusted for age, sex, diagnosis, revisions, and BMI). After 1 year, the cemented tibial components migrated 0.043 mm/year, the HA-coated tibial components migrated 0.057mm/year, and the uncoated tibial components migrated 0.067 mm/year (p = 0.003, GLMM adjusted for age, sex, diagnosis, revisions, and BMI).

Migration from 1 to 16 years for individual cases with 10 years or more of RSA follow-up is presented in Figure 4 according to fixation type: HA-coated (8 cases), uncoated (9 cases), and cemented (9 cases). There was 1 knee in the HA-coated group, 1 knee in the uncoated group and 1 knee in the cemented group with continuous migration. The patient in the HA-group and cemented group has died with the TKA in situ. The patient in the uncoated group is still alive and is considered to be at risk of aseptic loosening.

At 10 years postoperatively, there was a mean increase in KSS compared to preoperatively (59 (CI: 54–66)). There were no statistically significant or clinically relevant differences in KSS between the fixation types (p = 0.9, GLMM adjusted for age, sex, diagnosis, revisions, and BMI) (Table 2).

At 10 years postoperatively, there was a mean increase in KSS function compared to preoperatively (33 (CI: 21–46)). There were no statistically significant or clinically relevant differences in KSS function between the fixation types (p = 0.4, GLMM adjusted for age, gender, diagnosis, revisions, and BMI) (Table 2). There were no significant differences in HSS or flexion between the fixation types.

The FTA angles were similar between the fixation types (p = 0.3, GLMM adjusted for age, sex, diagnosis, revisions, and BMI) (Table 2). There were no statistically significant or clinically relevant differences in alfa, beta, or delta angles between the fixation types. Ten years postoperatively, there were 2 partial 2-mm radiolucent lines in the HA group, 1 partial 2-mm radiolucent line in the uncoated group, and no radiolucent lines of 2 mm or more in the cemented group.

7 knees were revised: 3 tibial components for aseptic loosening (2 uncoated and 1 cemented), 3 for septic loosening (2 uncoated and 1 cemented), and 1 for instability (HA-coated). The individual migration patterns are shown in Figure 5. Of note is 1 case that was revised for secondary aseptic loosening after a period of stability. This secondary loosening was due to scalloping osteolysis at the tibial component located anterolaterally as identified on CT-scan and during the revision procedure. There was 1 case that was revised after late infection. This case also showed increasing migration after a stable period. There was 1 case of wound necrosis (cemented tibial component) early postoperatively that was treated successfully with surgical debridement and antibiotics, so the prosthesis was preserved.

There were 7 PE insert exchanges for wear: 2 in the cemented group, 3 in the HA group, and 2 in the uncoated group. There was no statistically difference between the groups in the rate of insert exchange (with the numbers available; HA-coated vs. cemented, HR = 1.0, CI: 0.2–6.9; p = 1.0; and HA-coated vs. uncoated, HR = 0.6, CI: 0.1–5.6; p = 0.6).