We evaluated 24 retrieved tibial components obtained from our implant retrieval laboratory. We chose implants that were in situ for more than 12 months from a single manufacturer that was able to provide matching new liners from their existing inventory. Selection of the implants was further influenced by the process of matching on patient sex, age, height, weight, revision diagnosis, and time in vivo. The selection process described was chosen to limit differences between manufacturers and patients allowing a better comparison of fixed-bearing and rotating-platform designs. The components included 12 rotating-platform and 12 fixed-bearing inserts (Table ). New polyethylene inserts were obtained from the manufacturer based on product numbers. All inserts were cleaned and stored at room temperature. Twenty-four control and 24 retrieved tibial polyethylene inserts were scanned. Funding for the study was provided by DePuy (Warsaw, IN, USA). Institutional review board approval was obtained because protected patient data were used. All implants used for this study were FDA-approved.
The inserts were scanned at 80 kVp and 450 μA with an isotropic voxel resolution of 92 μm (0.0036 inches) and a 1024 × 1024 in-plane image matrix at the Cleveland Clinic (ImageIQ, Cleveland, OH, USA) using a high-resolution micro-CT scanner (GE eXplore Locus; GE Healthcare, Waukesha, WI, USA). When the mediolateral dimension of an insert exceeded 9.4 cm, serial scans were obtained and subsequently stitched together to reconstruct the entire specimen geometry. Microview Image Analysis Software Version 2.2 (GE Healthcare) was used to filter the reconstructed image data using a 3 × 3 × 3 Gaussian smoothing filter. The DICOM-formatted image data were then imported into three-dimensional image analysis software developed by the Mayo Clinic’s Biomedical Imaging Resource (Analyze 9.0; AnalyzeDirect, Inc, Overland Park, KS, USA) that was used to reconstruct the three-dimensional insert geometry. Although the Gaussian smoothing filter is useful for suppressing normal image noise, it cannot eliminate image artifacts associated with metal pins that were embedded in the Low Contact Stress (DePuy) rotating-platform retrieved inserts. When present, these residual image artifacts were manually removed from each image slice.
Three-dimensional geometries for the retrieved and control inserts were created by segmenting the polyethylene from the background using a threshold Hounsfield value that defined each voxel in the image as either polyethylene or background air (Fig. ). The threshold value used for each image was chosen so that the volume of the polyethylene from the micro-CT image data matched the volume of the actual insert calculated by dividing the weight of the specimen by the nominal density of the polyethylene. A nominal density of 935 kg/m3
was used for extruded GUR 1020 inserts and 931 kg/m3
was used for extruded GUR 1050 inserts. The accuracy and error associated with this imaging technique have been previously reported [21
Fig. 1 Three-dimensional rendered image of a control insert (left) and retrieved insert (right). Notice the visible implant identifiers on the backside of the control (lower left). Likewise, the absence of identifiers and the unworn central tray hole on the (more ...)
Using the Analyze software, one of the authors (RLZ) who was not a surgeon manually superimposed the worn and control insert geometries at the unworn intercondylar eminence on the femoral articulating side of both liners. The superimposed images were visually inspected to check for the conformity along other surfaces including the peripheral (lateral) edges and backside surface. To account for the possible influence of manufacturing tolerances and in vivo polyethylene expansion resulting from fluid absorption, if the peripheral edges of the control insert were uniformly outside or inside the boundaries of the retrieved insert, the control insert was dilated or eroded to match the geometry of the retrieved insert. Additionally, in three cases in which the backside of the control insert was uniformly larger or smaller than the retrieved insert and the machining marks on the backside surface of the retrieved insert were present or the uniform geometrical differences between the control and retrieval extended across contacting and noncontacting portions of the backside surface (such as chamfers), the backside surface of the processed control was uniformly dilated or eroded by a single slice (92 μm) to match the retrieved insert backside surface. The net change from this process resulted in 15 cases in which the control was dilated because it was slightly smaller than the retrieval and seven cases in which the control was eroded because it was slightly larger than the retrieval. Although we might erode or dilate the lateral edges and backside surfaces, the topside articular surface was never modified. The mean change in volume from this process was 0.4% ± 0.5% (range, 0.2% erosion to 1.1% dilation) of the control volume or 136 ± 142 mm3. Without this adjustment, 15 cases would have had decreased wear and seven would have had inflated wear.
We defined penetration as the volume of polyethylene from the control insert that is outside the boundary of the retrieved insert. Deformation is defined as the volume of polyethylene from the retrieved insert that is outside the boundaries of the control. It follows that wear (material lost from the control) is defined as penetration minus deformation (Fig. ).
Definitions of penetration, deformation, and wear are shown.
To validate the technique, physical measurements of the polyethylene liners were compared with measurements made with the micro-CT technique described. The thickness of each retrieved and control insert was measured by sampling the surface until the thinnest location in both the medial and lateral compartment was determined and measured with digital calipers (Mitoyo, Tokyo, Japan). The location of the measurement was recorded so that the same location could be measured on the micro-CT scan. We observed a very strong correlation (r2
= 0.998, p < 0.001) between all of the caliper and micro-CT thickness measurements (Fig. ). The Bland and Altman Limits of Agreement [3
], corresponding to an interval containing 95% of the differences between the physical caliper and micro-CT image measurements, ranged from −0.16 mm to 0.16 mm.
Fig. 3 A strong correlation (p < 0.001) was found comparing insert thickness measured with calipers versus insert thickness based on three-dimensional reconstructions derived from micro-CT images. The dashed gray line represents perfect (more ...)
To apply the technique, we compared the total penetration, deformation, and wear values for the rotating-platform and fixed-bearing implants. Each registered three-dimensional image was divided into topside and backside regions to allow rotating-platform to fixed-bearing comparisons. Rates were determined by dividing volumetric measurements by time in vivo.
Linear regression was used to evaluate the relationship between the raw wear volumes (computed without accounting for manufacturing tolerances) and the adjusted wear volumes. Owing to the nonparametric distribution of wear data, comparisons of rotating-platform to fixed-bearing penetration, deformation, wear, and comparisons of topside and backside wear were evaluated using a Mann-Whitney U test. All statistical analyses were performed using SPSS (Statistical Package for the Social Sciences, Chicago, IL, USA).