Construct Damage and Loosening Around Glenoid Implants: A Longitudinal Micro-CT Study of Five Cadaver Specimens

Abstract

The evolution of failure of bone and cement leading to loosening of glenoid components following shoulder arthroplasty is not well understood. The purpose of this study was to identify and visualize potential mechanisms of mechanical failure within cadavers, cemented with two types of components, and subject to cyclic loading. Five glenoid cadaver bones were implanted with either a three-pegged polyethylene component, or prototype posteriorly augmented component which addresses posterior bone loss. Specimens were loaded by constant glenohumeral compression combined with cyclic anterior-posterior displacement of the humeral head relative to the glenoid. At six time points across 100,000 cycles, implant loosening micromotions were optically measured, and specimens were imaged by micro-computed tomography. Scans were 3D registered and inspected for crack initiation and progression, and micro-CT based time-lapse movies were created. Cement cracking initiated at stress concentrations and progressed with additional cyclic loading. Failure planes within trabecular bone and the bone-cement interface were identified in four of the five specimens. Implant subsidence increased to greater than 1.0 mm in two specimens. Cemented glenoid structural failure can occur within the cement, along planes of trabecular bone, or at the bone cement interface.

Introduction

The annual number of shoulder arthroplasties in the United States increased 2.5-fold between 2000 and 2008.¹ Glenoid loosening has been reported as the most common complication in total shoulder arthroplasty.² Although cemented implant loosening may often involve biological processes such as osteolysis,³ loosening and migration is also ultimately a mechanical failure of the underlying complex structure⁴ consisting of interlocking cement and trabecular bone. Glenohumeral joint forces can often reach full bodyweight,⁵ and these forces are repeatedly transmitted to the glenoid eccentrically in the superior-inferior and anterior-posterior directions.⁶ Clinical loosening has been reported to occur within the bone or bone-cement interface, within the cement itself, or at the implant-cement interface,^3,7 but additional benchtop investigation is needed to reveal the mechanics of failure initiation and progression.

A common challenge in total shoulder replacement is that the native glenoid bone is worn posteriorly.⁸ In such cases the surgeon often either implants a standard component malaligned relative to the scapula, or reams away healthy bone on the anterior side to achieve normal alignment. In the case of a malaligned glenoid there is concern for posterior instability of the shoulder⁹ and early failure of the component, whereas in the case of anterior reaming, there is concern regarding taking away too much bone and shortening the vault, potentially precluding one’s ability to implant the component.¹⁰ Previously in our laboratory we developed a posterior-augmented polyethylene implant to compensate for this bone loss,¹¹ but more understanding is needed of the effects of posterior augmentation on potential glenoid loosening.¹²

Many previous studies have investigated loosening of different glenoid implant designs in the laboratory by a combined compressive and shear cyclic loading protocol.^12–17 Nearly all of these studies measured implant loosening but did not assess damage of the underlying structure during the loosening process, and used synthetic bone substitutes instead of human bone. Gregory et al.¹⁵ reported a study of six cadavers (and six synthetic specimens) in which keeled glenoid implants were cyclically loaded and evaluated by longitudinal clinical computed tomography (CT) scans, and microscopy at the end of loading. Their CT-detected implant-cement failure locations matched locations determined by microscopy well in regions near the glenoid face, but not deeper within the glenoid vault.

In the present exploratory study of five cadaver specimens we sought to identify and characterize potential mechanisms of mechanical glenoid loosening for standard pegged and posteriorly augmented glenoid implants. We hypothesized that underlying structural damage could develop during cyclic mechanical loading at various locations across the implant/cement/bone interface, not only primarily at the implant-cement interface as previously reported.¹⁵ We imaged specimens using longitudinal high resolution micro-CT performed at different time points of cyclic loading, then performed 3D image registration and visual image inspection. In addition we measured gross implant loosening at the same time points as the micro-CT, using 2D marker tracking from photographs of the implant rim under static loads. This approach enabled, for the first time, detailed assessment of the time-course of gross glenoid component motions and underlying structural damage in implanted, cyclically loaded cadavers.

Methods

Specimens

Five glenoid bone specimens (Table 1) were isolated from four fresh-frozen cadavers. An additional two specimens were also dissected but not ultimately used; one was a pilot specimen (implanted with a standard component) in which we used a different potting method which led to gross cortex failure, and in another one (implanted with a posterior augmented component) the glenoid vault cortex was perforated during component implantation. The specimens did not have any apparent previous trauma or pathologic abnormalities, including arthritis or glenoid face deformity. Each glenoid was stripped of soft tissues and isolated from the scapula by osteotomy 40 mm medial to the glenoid face, and the acromion and coracoid were removed. Specimens were encased in 50 x 50 x 40 mm blocks of polymethylmethacrylate (Ortho-Jet BCA, Lang Dental, Wheeling IL). The potting was performed with the glenoid mounted in a temporary aluminum box such that the glenoid face plane and superoinferior direction were perpendicular to the sides of the box. Components were implanted with their articular surfaces parallel to the glenoid face plane (in the specimen’s native version angle) as follows.

Table 1.

Cadaver specimen data.

Specimen Label	Implant Type	Donor Gender	Donor Age of Death	Arm Side
S1	3-pegged commercial	female	not available	right
S2	3-pegged commercial	female	71	left
PA1	posterior augmented	female	71	right
PA2	posterior augmented	female	70	right
PA3	posterior augmented	male	not available	right

Note: PA1 and S2 were paired from same donor.

Two specimens (S1 & S2, Table 1) were implanted with commercial 3-pegged, all polyethylene, 46 mm diameter surface glenoid components (Bigliani/Flatow, Zimmer, Warsaw IN). The radiopaque metal pin in the central peg was removed (to eliminate later micro-CT artefact) using a sharpened thin-walled metal tube, and an acetal pin was inserted in its place. Specimens were prepared using symmetric reaming of cartilage with minimal violation of the subchondral bone, and implants were cemented (Surgical Simplex P, Stryker, Kalamazoo MI) by an experienced shoulder surgeon using the implant manufacturer’s recommended technique.

Three additional specimens (PA1, PA2, & PA3) were implanted with posterior augmented glenoid components (Fig. 1). Each component was constructed by fastening a 5 mm thick polyethylene block to the posterior-side back of the same type of commercial component described above. The block was computer-control-milled so that its surface conformed with the pegged component, and block fixation was obtained using special-purpose epoxy (Bondit B-45TH, Reltek, Santa Rosa CA) and a doweled mechanical interlock. Cadaver specimens were prepared and implants cemented using standard technique, except a 5 mm deep bone block was removed posteriorly to mate with the posterior augmented implant as previously described.¹¹

Figure 1.

Preparation of the posterior augmented glenoid component and specimen: (A & B) The component was created by rigidly fastening a conforming 5-mm thick block to the posterior back portion of a 3-pegged component. (C) The glenoid bone was prepared by cutting a step into the posterior side with custom tooling. (D) Final implanted specimen, potted in a block of PMMA.

Mechanical loading

Specimens were mechanically loaded cyclically in the anterior-posterior direction using a method based on an ASTM protocol for assessing glenoid loosening.¹⁸ Each specimen was fastened to a carriage with the glenoid implant anteroposterior direction pointed vertically, and the carriage rode on a horizontal linear bearing (Fig. 2). Separately, the commercial humeral head component matching the 46 mm glenoid was welded to a vertical linear bearing, and the humeral head position was controlled by a servohydraulic test frame (Interlaken 3300 with Flextest 40 controller, MTS, Eden Prairie MN). The center position of the humeral head relative to each glenoid was established and confirmed by cycling under low load and displacement.

Figure 2.

Cyclic loading apparatus: (A) Schematic showing how +/− 2.5 mm displacement was applied by the humeral head to the glenoid in anterior-posterior direction, while a constant 750 N glenohumeral compression was also applied. The camera was used to measure implant toggle and subsidence. (B) Photograph of the apparatus. (C) Close up of humeral head and glenoid specimen. The implanted cadaver bone can be seen potted in a block of acrylic cement.

A 750 N glenohumeral compressive force was applied by a low friction pneumatic cylinder connected to the specimen carriage and an in-line load cell (Fig. 2A). While under compression the humeral head was translated +/− 2.5 mm in the anterior-posterior directions, with frequency of 1 Hz, relative to the center of the glenoid. This loading model is based on an ASTM standard.¹⁸ The compressive force magnitude is supported by previous in vivo instrumented implant studies which showed glenohumeral joint forces reaching, and sometimes exceeding, bodyweight.⁵ Anterior-posterior cyclic displacement of the humeral head is also supported by a previous in vivo study.⁶ The frequency of 1 Hz is believed to be physiologically relevant, but was also chosen because of practical considerations including proper experimental loading control. The displacement magnitude of +/− 2.5 mm was approximately 90% of the subluxation distance according to pilot testing of the same implant type.^15,18 A total of 100,000 cycles of humeral head displacement, representing 25 high-load shoulder movements per day for ten years,^5,13 were applied to each specimen.

Micro-CT and image analysis

At 0, 1000, 10000, 25000, 50000, and 100000 cycles, specimens were scanned by micro-CT (vivaCT 40, Scanco Medical, Brüttisellen Switzerland, energy 55kV/145μA). Images were reconstructed as 2k x 2k matrices of 19 μm isometric voxels. Images were then filtered and resampled to 38 μm cubic voxels using Matlab (Mathworks, Natick MA).

For each scan, coordinates of four landmarks on the implant, including the three tips of the pegs and a second point on the central peg, were identified in Avizo software (VSG US, Burlington MA). Image volumes from each time point were three-dimensionally co-registered with the 0-cycle scan using a two-stage process with routines available in Avizo: (1) using a landmark based least-squares method; followed by (2) intensity-based registration using a mutual information metric and extensive direction optimizer. Image volumes were then visually inspected in all three orthogonal planes. Because the scans were co-registered together, damage evolution due to loading could be detected by comparing each scan to the scans from previous time points. The cement mantle around each peg was inspected, and cracks were categorized as either through-mantle (crossing from implant to bone space) or partial-mantle.

Time-lapse movies at coronal and transverse planes passing through the central peg were also created from co-registered images at these planes. These movies were carefully reviewed in order to identify failure locations within the trabecular bone and at the bone-cement interface. The movies were created for only two orthogonal planes, but were useful for identifying subtle damage that was not detected when inspecting the images themselves.

Implant toggle and subsidence measurements

Prior to loading, small marks were made on anterior and posterior implant rims and on adjacent bone potting (Fig. 3). At 0, 1000, 10000, 25000, 50000, and 100000 cycles, cyclic loading was halted. A 15 megapixel digital single-lens reflex camera with macro lens recorded coronal plane images via a mirror mounted above the glenoid. Photographs of a ruler were used to establish the scale, which was approximately 6 μm/pixel. Photographs were recorded of the anterior implant rim under the following three glenohumeral static loading conditions: (1) central loading with the humeral head at the glenoid center; anterior loading with the humeral head displaced 2.5 mm anteriorly, and posterior loading with the humeral head displaced 2.5 mm posteriorly. In each loading condition 750 N mediolateral force was applied to the humeral head.

Figure 3.

Example images used to determine toggle and subsidence: (A) Uncropped high resolution photograph showing (from top to bottom) PMMA potting, edge of glenoid bone, implant anterior aspect, and metal humeral head. (B) Close-up of the dashed area. For each time point, the distance was determined between two landmarks which could consistently be identified on the glenoid and potting.

These coronal plane, high resolution photographs of the anterior implant rim were imported into Image J (National Institutes of Health). Landmarks on the implant and potting were identified that could be readily seen across all images for the specimen (Fig. 3). Implant-bone distances between these landmarks were measured. Subsidence was defined for the central glenohumeral loading condition as the change in implant-bone mediolateral distance relative to the 0 cycle time point. Toggle was defined as the difference in implant-bone distance between the anterior and central loading conditions, or between the posterior and central loading conditions. Half of the toggle measurements were repeated on the same images by a second observer, and absolute interobserver differences averaged 0.030 mm.

Results

Micro-CT analysis revealed several different types of internal failure of the implanted cadaveric glenoid constructs, indicated by visible cracking, or relative movement of structures from one time point to the next:

• gross medial shifting of an implant peg and its surrounding cement mantle, in Specimens S2 & PA1 (Fig. 4, Supplementary movies)

• progression of cracking through cement (Figs. 5 & 6).

• failure of trabeculae surrounding the cement mantle, or along planes located in the lateral half of the glenoid vault (Figs. 4–6, Supplementary movies)

• motion between trabeculae and apposing cement, i.e. failure of the bone-cement interface (Figs. 4 & 6, Supplementary movies)

• small gapping at implant-cement interface, often at the glenoid face but in some cases deep within the glenoid vault (Fig. 6: e.g. Spec. S2)

Figure 4.

Locations of trabecular bone and bone-cement interfacial failure (orange curves) in a coronal plane (left) and transverse plane (right) passing through the central glenoid component peg. Failure locations were determined from the corresponding time-lapsed micro-CT movies (Supplementary Material). The images from before loading (0 cycle) are shown here. In the images, bone appears whiter than the cement. Arrows indicate locations and directions of observed implant deformation. S1 & S2 were specimens implanted with standard components, and PA1, PA2, & PA3 were specimens implanted with posterior augmented components.

Figure 5.

Progression of cement and bone cracking (red arrows) seen from co-registered micro-CT images of a posterior-augment implanted specimen. The added posterior step of the implant is visible in the bottom-right of the images.

Figure 6.

Representative images of cement, bone, cement-bone interface, and cement-implant interface failure (all circled in orange) from the five implanted and cyclically loaded cadaver specimens. The cycle count (‘K’ = 1000), and which of the three pegs the damage occurred around, are provided for each image.

In the cement of these specimens, cracking tended to increase with cycling (Fig. 5, Fig. 7C). After 100,000 cycles the posterior augment implant specimens showed cracking in either two or three cement mantles surrounding the three pegs, whereas the standard implanted specimens showed cracking in only one (Fig. 7C). Generally these cracks first appeared small at stress concentrations, then progressed to larger cracks. Cement cracks could often be traced to stress concentrations such as the grooves in the pegs (Fig. 6: e.g. Spec. S2) and the corner of the posterior augmentation (Fig. 5). Cement cracks occurred more at locations surrounding the lateral half of the pegs (nearer to the glenoid face) in these specimens.

Figure 7.

Measurements of component micromotions (A & B) based on the external high resolution photographs, and characterizations of cement cracking (C) based on 3D-registered micro-CT images, in all specimens and all time points. (A) Toggle micromotion is of the anterior component rim during static posterior eccentric loading, and (B) subsidence is determined during static central loading.

Specimen S1 (implanted with a standard component) exhibited the least amount of damage, and it was noted that this specimen had the most congruent seating of the component back onto the glenoid face (Fig. 4). All three posterior augmented specimens exhibited trabecular failure in the region beneath the posterior augment block (Fig. 4, Supplementary movies).

None of the implants exhibited gross loosening after the full 100,000 cycles. Most toggle micromotions increased in magnitude due to the cycling, but in the majority of tests, greater than 75% of the final 100,000 cycle-micromotion was already present at 0 cycles (Fig. 7A, Fig. S-1). Toggle of the anterior implant rim during posterior loading across all five specimens averaged 0.56 mm (± 0.22 mm, range 0.22–0.82 mm) at the start of cyclic loading, and averaged 0.66 mm (± 0.12 mm, range 0.51–0.81 mm) after 100,000 cycles of loading. Subsidence of the implant into the bone, conversely, tended to have larger increases during cycling in these specimens (Fig. 7B). After 100,000 cycles loading, subsidence for all specimens averaged 0.76 mm (± 0.50 mm, range 0.12–1.44 mm), and exceeded 1.0 mm in one standard and in one posterior augment implanted specimen. Also after 100,000 cycles of loading, the specimens had an average of 2 (± 1, range 1–3) of the 3 cement mantles with partial or through-cracking according to the micro-CT image analysis (Fig. 7C).

Discussion

This cadaveric study is the first, to the best of our knowledge, to longitudinally assess internal construct micro-damage in implanted cadaveric glenoids. Cement cracking initiated at stress concentrations surrounding the upper half of the pegs during the first 10,000 cycles in most specimens, and progressed across the cement mantle with additional cyclic loading. Failure also occurred within planes of trabecular bone or at the bone cement interface in some specimens (Figs. 4 & 6, Supplementary material movies). This study is novel in its visualization and characterization of potential mechanisms of glenoid mechanical loosening in cadaver specimens including data at the sub-50 micron scale. Hypothesis testing was not performed with our small sample sizes; instead our study should be viewed as exploratory with the purpose of identifying potential mechanisms of loosening enabled by using a novel image analysis approach.

Our results are somewhat in contrast to a previous study which used longitudinal clinical CT of glenoid cadavers implanted with keeled components and subjected to cyclic superoinferior loading.¹⁵ That study showed failure of fixation occurred only between the polyethylene implant and cement, not within the bone or bone-cement interface as was observed in our study. Differences between our results and theirs may be attributable to differences between the studies in loading, implant type (keeled vs. pegged), and image analysis methods. Our finding of failure in trabecular bone immediately surrounding cement mantles is in general agreement with a recent micro-finite element modeling study from our lab which showed high stresses in these regions.¹⁹

Most previous studies of glenoid loosening used rigid synthetic bone substitute for pragmatic considerations, and did not directly assess the internal implant-cement-bone structure.^12,14,16 The purpose of those studies was generally to assess gross loosening for particular component designs, not to identify the mechanisms of loosening. Bone cement does not spread into rigid synthetic bone, which is highly nonporous, as it does into real glenoid trabecular bone. In pilot testing we attempted to use an alternative porous synthetic bone (1522-11 cellular rigid polyurethane foam, Sawbones, Vashon WA), but cement also penetrated differently into this synthetic compared to real bone as observed in micro-CT images (Fig. S-2), and the synthetic bone was weaker than real bone. These initial findings led us to perform the present experiment using cadaver bone.

This study gives preliminary indication that subsidence may be a better indicator of internal construct damage than toggle micromotion (Fig. 7), although the latter is emphasized in current standards for testing of glenoid implant loosening. Clinically, subsidence or tilting of the glenoid have been used as indications for failure or “at risk” of failure for glenoid components.^20,21 The reported rates of subsidence or tilting in clinical outcome studies were up to 10 % at 5 years and 33% at 9 years.^20,21

This study also provides initial indication of the mechanics of loosening of posterior augmented implants, and suggests design changes that might mitigate the risk of mechanical failure. Traditionally, large posterior defects in osteoarthritic shoulders requiring glenoid replacement have been treated surgically with a posterior bone graft. The first clinically used augmented glenoid component used a wedge-shaped posterior augment.²² Unfortunately, this implant showed early failure bringing into question whether augmented glenoids were feasible to deal with posterior bone loss. In that study, however, optimal alignment of the component could not be verified so it brings into question whether it was the actual implant that was not feasible or whether the component was implanted in a residual retroverted alignment which could compromise its performance. Patients with large posterior defects remain very difficult cases in shoulder replacement. The posterior augmented implant has the important advantage of restoring proper alignment with limited removal of healthy bone in these patients. We recently reported on the glenoid cortical strains following step glenoid component implantation in a biomechanical cadaveric study with simulated muscle loadings.¹¹ Differences were not detected in periglenoid strains between specimens implanted with polyethylene-step augmented glenoid components compared to those implanted with standard components. That initial benchtop investigation supported that a properly aligned posterior augmented glenoid component may be a feasible option to deal with posterior glenoid erosion. In the current investigation with limited sample sizes, toggle micromotions and subsidence were not markedly different between the standard and posteriorly augmented glenoids (although statistical comparisons were not performed), and there were no gross failures of either implant type. Our study provides indication that bone failure can occur immediately medial to the posterior augment, although this finding may be an effect of the increased reaming and violation of subchondral bone required to create the 5 mm posterior step defect in our nonarthritic cadaveric glenoids compared to the reaming typically performed clinically. Furthermore the corner of the augment block can be a stress riser for crack initiation. With design modifications of the step glenoid component, such as filleting the posteromedial edge of the augmentation, it may be possible that the rate of cement damage could be decreased, and ultimately a stepped posterior augmented implant could potentially have a lower clinical complication rate than a posterior bone graft reconstruction.

A primary study limitation is the limited number of samples successfully tested. An additional two pilot specimens were not ultimately used because of either cortex fatigue failure after using a different potting method, or glenoid vault cortex perforation during implantation. The study is also limited by the use of glenoids from donors without arthritis. Arthritic glenoids often have articular surface deformities such as posterior wear and abnormal version angle, and potential subchondral and trabecular bone architecture changes not captured by our cadaver model. Such pathologic changes impact the amount of glenoid reaming and whether symmetric or asymmetric reaming is performed in preparing the glenoid, both of which are significant factors in the mechanism of glenoid component loosening that could not be tested in the current study. Only one size of posterior augmentation block was tested. Due to the use of non-pathologic glenoids in this study, the specimen preparation needed to accommodate the posterior augmented glenoid component removed a large amount of subchondral and trabecular bone which may have substantially affected the bone failure that occurred in this area. Implantation of augmented step components in actual patients may aim to preserve at least part of the posterior sclerotic bone of the arthritic glenoid surface, potentially providing more mechanical support to the implant than modeled in our study. As in other cadaveric studies, the study is limited by the lack of consideration of in vivo biological response of the bone over time. In a previous study of postmortem-retrieved femoral stems, it was shown that bone remodeling from in vivo service likely resulted in less cement-bone contact and a more compliant cement-bone interface.²³ Our specimen potting provided constraint around the glenoid vault during cyclic loading, similar to a previous study¹⁵, although this may have artificially stiffened the glenoid cortex compared to in vivo loading conditions. The constraint method was partially for pragmatic reasons; in pilot testing of a specimen potted more medially (not surrounding the glenoid vault), the vault cortical bone grossly cracked. Toggle micromotions were determined based on tracking the implant rim relative to the bone potting, not the bone itself. Anterior-posterior, not superior-inferior, cyclic loading was applied because we believed the former was a better test of the posterior augmented component; in vivo eccentric loads occur in both anterior-posterior and superior-inferior directions.^6,12

Supplementary Material

Supp FigS1. Figure S1.

Toggle micromotions of the anterior and posterior implant rims during anterior and posterior static loading, at 0 and 100,000 cycles.

Supp FigS2. Figure S2.

Micro-CT images of glenoid components cemented in cadaver bone and synthetic bone.

specimenPA1_coronal. Videos.

Time-lapse movies created from micro-CT datasets co-registered from different timepoints. For each specimen implanted with either a posterior augmented component (PA1, PA2, and PA3), or standard component (S1 and S2), videos were created from coronal and transverse cross-sections passing through the central glenoid component peg, with filenames listed below indicating the corresponding specimen and cross-section view. Markings indicate visible regions of damage.