|Home | About | Journals | Submit | Contact Us | Français|
This study examined whether patients with rheumatoid arthritis (RA) demonstrate different patterns of prosthetic wear or cellular responses to implant wear debris compared to patients without inflammatory joint disease.
Thirty-eight patients who had a primary revision of a total elbow arthroplasty (TEA) for aseptic loosening between 1996 and 2008 were identified. Twenty-five had RA and 13 had no inflammatory arthritis. Clinical data, gross wear patterns of the removed prostheses, and histopathological analyses of peri-implant tissue were compared between RA and non-RA patients.
Evaluation of the retrieved prostheses showed that conformational change of the humeral polyethylene bushing was associated with the generation of polyethylene and metal particles. The amount and type of wear debris in peri-prosthetic tissues was similar in RA and non-RA patients. RA patients not on anti-tumor necrosis factor (TNF) therapy exhibited a histologic pattern of interstitial and sheet-like lymphocytic infiltrates associated with a high plasma cell composition, which was different from the predominantly perivascular infiltrates with few plasma cells seen in non-RA patients (p-value = 0.04). RA patients on anti-TNF therapy fell in between these two groups.
RA patients exhibit a distinct cellular response to implant wear debris compared with non-RA patients. This reaction was unrelated to differences in the type or amount of wear debris and was mitigated by anti-TNF therapy. These results suggest an intrinsic alteration in immunoregulation in RA and have implications for potential immunologic treatment of osteolysis in these patients.
Peri-implant osteolysis and associated loss of fixation is the major reason total joint replacements fail. The progressive loss of bone adjacent to the implant has been attributed to a granulomatous inflammatory reaction induced by particulate implant wear debris at the bone-implant interface . Patients frequently exhibit dramatic differences in the rate of peri-implant bone loss, and it is unclear whether this reflects differences in the properties or amount of wear debris, differential patterns of biomechanical failure, or differences in individual host immune response.
Studies by Caplan and coworkers noted that coal miners with RA who were exposed to silica containing coal dust were uniquely prone to the development of large pulmonary granulomas (“Caplan’s nodules”) . They speculated that this differential response to an inorganic particulate reflected an intrinsic alteration in immunoregulation among RA patients. Similarly, RA patients may exhibit a unique immune response to the inorganic wear particles generated by prosthetic implants, which might lead to different failure mechanisms compared with non-RA patients. Kaufman et al. reported on a series of patients undergoing revision surgery for aseptic failure after TEA . They did not note a difference in the cellular features of the peri-implant tissue reaction in patients with or without inflammatory arthritis. More recently, Goldberg et. al. evaluated the tissue reaction accompanying failed, semi-constrained, Coonrad-Morrey elbow implants at primary revision in sixteen patients, nine of whom had inflammatory arthritis . Similarly, the peri-prosthetic tissue showed no evidence of a differing cellular reaction on the basis of underlying diagnosis. In contrast, Goldring found that among a similar group of patients, lymphoplasmacytic inflammation was found exclusively in RA patients . Factors responsible for these contradictory findings could include differences in the composition and types of devices, differences in patient populations, and differences in treatment. The present study was undertaken to compare cellular features of the peri-implant tissue response among patients with and without RA undergoing revision of a TEA due to aseptic loosening, and to ascertain the relationship with underlying disease, patterns of gross device wear, and the amount and type of particulate debris.
Surgical specimens from all primary TEA revisions performed between 1996 and 2008 for aseptic loosening were identified from the database of the Department of Pathology and Laboratory Medicine at the Hospital for Special Surgery (HSS). Implant components corresponding to the identified cases were recovered through the Implant Retrieval Registry in the Hospital’s Department of Biomechanics. The patients’ primary diagnosis, gender, race, age at primary TEA, age at revision TEA, pre-revision medical treatment, pre-revision flexion-extension range of motion, dominant hand, as well as height and weight at the time of revision were obtained from medical records. Non-RA patients include non-inflammatory arthritis patients with a history of OA or trauma. RA patients were labeled treated or untreated based on pre-operative usage of a TNF-inhibitor. When stratifying patients on this basis, those in whom this data was unavailable were excluded from analysis. Duration of implantation was defined as the number of years between the original TEA and primary revision TEA. A short duration of implantation was <4 yrs; a long duration of implantation was ≥ 4 yrs.
HSS Institutional Review Board approval was obtained for this study.
As available, the retrieved humeral and ulnar components, humeral polyethylene bushing (Fig. 1A), and axle were evaluated for evidence of wear and surface damage. Due to the limited number of metal components (Fig. 1B), no information was collected on metal loss.
A previously developed subjective scoring method was used to assess seven modes of polyethylene damage . Of the seven modes, conformational change, due either to deformation or material loss, was considered the most accurate and functionally relevant metric of biomechanical damage. Grades of none, mild, moderate or severe corresponded with 0, <10%, 10–50%, and >50% of the articulating surface area being visually thinner than the non-articulating surface. However, this scoring system was insufficient to estimate the extent of wear in instances of penetrating focal damage. Focal wear was measured separately as the proportion of thinning in a localized area between the inner and outer bushing diameters compared to a non-deformed region. Focal wear was graded as none, mild, moderate, or severe corresponding to the diameter being 0, <10%, 10–50%, and >50% thinner compared with the non-deformed surface. A device’s final grade was the highest qualifying score on the basis of either surface area damaged or focal wear.
All grading was done under a stereomicroscope while blinded to both the underlying diagnosis and tissue histology. Grading was performed independently by two readers (AV and DC). Differences were resolved by consensus.
Hematoxylin and eosin stained samples of peri-prosthetic tissue were evaluated for the type and size of particulate material, as well as the presence, type, location, and intensity of inflammatory infiltrates. This analysis is based on a modification of a similar analysis published elsewhere .
Polyethylene debris particles were classified as either small (<500µm) or large (≥500 µm) based on the largest dimension (Fig. 1C). Metal particles were not stratified by size, as they were uniformly small particles (~10µm) with little heterogeneity on the basis of visual identification. The distribution of polyethylene debris was defined as low prevalence or high prevalence. Low prevalence was defined as wear particles seen in ≤ 50% of the area of at least 2 histologic fields viewed at 10× magnification. For metallic debris, which was less common, low prevalence was defined as particles seen in ≤10% of the area of at least 2 histologic fields viewed at 10× magnification.
In describing the peri-prosthetic tissue reaction, monocytes and giant cells were defined as low prevalence if they comprised ≤ 50% of the cellular infiltrate on two separate 10× fields, and high prevalence if they comprised >50%. A lymphocytic aggregate was defined as a group of ≥10 lymphocytes visualized on a 5× field. Perivascular infiltrates surrounded a discernable vascular structure (Fig. 1D). Focal interstitial infiltrates were defined as aggregates of lymphocytes in the absence of a discernable vascular structure (Fig. 1E). Sheet-like aggregates were defined as a diffuse interstitial pattern of lymphocytic infiltrates throughout the entire tissue section (Fig. 1F). Focal interstitial and sheet-like aggregates represent two degrees of intensity of interstitial lymphocytic inflammation. When evaluating plasma cells within lymphocytic aggregates, low prevalence of plasma cells was defined as comprising ≤ 10% of the lymphocytic aggregate, and high prevalence >10% of the aggregate in two separate 10× fields.
To confirm that our histologic identification of cell types was correct, immunohistochemistry was performed on tissue from a representative subset of cases. T- and B-lymphocytes were identified by using anti-CD3 and anti-CD20 antibodies, respectively, and plasma cells by using anti-Kappa and anti-Lambda antibodies.
Available anterior-posterior (AP) and lateral pre-operative radiographs were evaluated for the presence of osteolysis based on a previously used scoring system . The humeral and ulnar stems of the prosthetic were divided into 4 zones each and examined for the presence radiolucencies at both the cement-prosthesis and cement-bone interface.
Fisher’s Exact test was used to compare all categorical variables. T-tests were performed on continuous data. A kappa analysis was performed to establish inter- and intra-rater reliability of the qualitative histologic scoring.
Of 88 cases, 11 were excluded because of infection. Of the remaining 77 cases, 39 had inadequate tissue for review. Complete medical records were available in 36 of the remaining 38 patients; two had only the operative record and partial medical records.
Twenty-five of the 38 patients had RA (66%), 7 had OA (18%), and 6 had a history of traumatic elbow injury without underlying inflammatory arthritis (16%). Of the 25 RA patients, 24 had pre-operative medication data. Ten RA patients were on a TNF inhibitor pre-operatively (etanercept in all cases). The remaining 14 RA patients were not on a TNF inhibitor pre-operatively. None of the non-RA patients were pre-operatively on a TNF inhibitor. 4 patients (3 RA and 1 non-RA) underwent revision surgery prior to FDA approval of etanercept.
Of all RA patients, 91% were Caucasian, 84% were women, and 95% were right-handed compared to 100%, 87%, and 100% respectively in non-RA patients. The right elbow was the operative site in just under half of both groups. Thirteen percent of RA patients and 33% of non-RA patients had a BMI >30 (p = 0.19). The average pre-operative range of motion was 112° (0 to 210°) for RA patients and 109° (60 to 130°) for non-RA patients (p = 0.85). RA patients were younger at the time of original implant surgery (p=0.0003). The average duration of implantation was 8.9 yrs for all RA patients and 5.3 yrs for non-RA patients (p = 0.06) (Table 1).
Untreated RA patients had a longer time to revision than treated RA patients (10.9 vs. 5.8 yrs, p = 0.04) or non-RA patients (10.9 vs. 5.3 yrs, p = 0.01). There was no difference in the duration of implantation between non-RA patients and treated RA patients (5.8 vs. 5.3 yrs; p = 0.77).
Of the 38 patients in this study, 24 had semi-constrained retrieved implants available for analysis. 12 were manufactured by Osteonics, 7 were Coonrad-Morrey manufactured by Zimmer, 2 were Solar manufactured by Stryker, 2 were Triaxial manufactured by Johnson and Johnson, and 1 was a custom device. All devices had humeral polyethylene bushings. Metal components were available for 15, of which 3 contained both humeral and ulnar components, 7 contained only the humeral component, and 5 contained only the ulnar component. Five of the axle-containing devices had no axle available for analysis. All metal components were titanium.
Conformational changes in the 24 humeral polyethylene bushings were mild in 10, moderate in 3, and severe in 11. Moderate changes were qualitatively much closer to severe than mild, so moderate and severe categories were grouped for statistical analysis. No difference in severity of conformational changes were noted between Coonrad-Morrey implants and other designs (p = 0.08). More severe conformational change was associated with a longer duration of implantation of >4 years (p = 0.01). Conformational change showed no association with underlying diagnosis (RA vs. non-RA, p = 1.0), or with use of TNF inhibitors (p = 1.0) (Fig. 2A), regardless of whether the 3 instances of moderate changes were excluded.
Polyethylene and metal particles were observed in varying quantities in the peri-prosthetic tissue reaction. Increased polyethylene particle prevalence (p = 0.01) and size (p = 0.005) and increased metal particle prevalence (p=0.05) were all associated with more severe conformational change (Fig. 2B-D).
Increased polyethylene particle prevalence (p = 0.06) and increased polyethylene particle size (p = 0.03) were associated with duration of implantation > 4 yrs. Metal particle prevalence was not associated with duration of implantation (p = 0.15).
No difference was found in the prevalence or size of polyethylene particles among non-RA patients, treated RA patients, or untreated RA patients (p ≥ 0.65 for all comparisons) (Fig. 3A-B). Similarly, no difference existed in the prevalence of metal particles among these three groups (p ≥ 0.68) (Fig. 3C).
A higher prevalence of foreign body giant cells was associated with a higher prevalence of polyethylene particles (p < 0.001), larger polyethylene particles (p = 0.004), and higher prevalence of metal particles (p = 0.04). There were no significant differences in the prevalence of foreign body giant cells (p = 0.73) between non-RA patients and RA patients. This finding did not change when RA patients were stratified by use of anti-TNF therapy (p = 1.0).
No association was found between the prevalence of monocytes and the prevalence of polyethylene particles (p = 0.24), size of polyethylene particles (p = 0.25), or prevalence of metal particles (p = 1.0). There were no significant differences in the prevalence of monocytes (p = 0.12) between non-RA patients and RA patients. This finding did not change when RA patients were stratified by use of anti-TNF therapy (p = 1.0).
The presence of any lymphocytic aggregates was associated with lower metal particle prevalence (p = 0.005) but not with polyethylene particle prevalence (p = 0.14) or size (p = 0.32). An increased plasma cell prevalence within lymphocytic aggregates was associated with widespread metal particles (p = 0.049) but was not significantly associated with polyethylene prevalence (p = 0.36) or size (p = 0.40).
In the presence of any type of wear debris, untreated RA patients showed an interstitial pattern of lymphocytic aggregation as opposed to the predominantly perivascular pattern seen in non-RA patients (p = 0.04). Treated RA patients showed a mixed pattern of perivascular and interstitial lymphocytic infiltrates intermediate between non-RA patients and untreated RA patients (Fig. 4A). 50% of untreated RA patients also showed a sheet-like intensity of lymphocytic infiltrates compared to none of the non-RA patients (p = 0.09). Treated RA patients showed a mixed intensity of lymphocytic infiltrates with 22% being diffuse, sheet-like infiltrates.
A similar pattern of lymphocytic infiltration was seen when considering only patients with small or few wear particles of any type. The majority of these patients had ≤ 10% plasma cells in the lymphocytic aggregates, with no significant difference when comparing underlying diagnosis, treatment, or particle composition.
Among patients with a high prevalence of polyethylene debris, including small and large particles, all untreated RA patients showed lymphocytic infiltrates with an interstitial or diffuse sheet-like appearance, associated with a high prevalence of plasma cells. This pattern was present in only half of treated RA patients. Only one non-RA patient exhibited this pattern of polyethylene debris, and in this patient the lymphocytic aggregates were perivascular and associated with a low prevalence of plasma cells (p =0.33) (Fig. 4B). The cellular patterns were similar in tissue sections containing predominantly large non-phagocytosable polyethylene particles (Fig. 4C). Again, all untreated RA patients had a high prevalence of plasma cells in lymphocytic infiltrates compared with non-RA patients (p = 0.07). Only one-third of treated RA patients exhibited this pattern.
In the presence of a high prevalence of metal debris, untreated RA patients had exclusively lymphocytic aggregates with an interstitial or sheet-like appearance that were associated with a high plasma cell prevalence (Fig. 4D). Unlike the histologic findings associated with polyethylene, 40% of the non-RA patients had lymphocytic aggregates with high plasma cell prevalence; however, unlike the RA patients, the plasma cells were associated with focal interstitial or perivascular lymphocytic aggregates. Treated RA patients had a prevalence of plasma cells similar to the non-RA patients (p = 1.0).
A representative convenience sample of peri-prosthetic tissue from 2 non-RA cases, 3 treated RA cases, and 3 untreated RA cases was selected for immunohistochemical staining. T- and B-lymphocytes were identified by using anti-CD3 and anti-CD20 antibodies, respectively, and plasma cells by using anti-Kappa and anti-Lambda antibodies. Repeat evaluation for the presence, type, location, and intensity of inflammatory infiltrates using cells identified via immunohistochemistry showed identical results to the reviews based on visual identification of cell types.
Of the 38 patients in this study, 18 had lateral radiographs and 17 of these 18 had AP radiographs available for review. These 18 patients included 5 non-RA, 3 treated RA, and 10 untreated RA patients.
At the cement-prosthesis interface, none of the RA patients had radiolucencies, regardless of treatment. One of five non-RA patients showed radiolucencies. At the cement-bone interface, 9 of 18 patients showed radiolucencies, including 2 non-RA, 2 treated RA and 5 untreated RA patients. There was no association between the presence of radiolucencies and treatment with etanercept, among RA patients.(p = 1.0). Similarly, there was no difference in the presence of radiolucencies between non-RA and RA patients (p = 1.0).
7 of 24 devices and 10 of 38 slides were randomly selected for kappa analysis. The inter-rater kappa values were 0.84, 1.0, 0.58, 0.62, 0.29, and 1.0, and the intra-rater analysis values were 0.63, 1.0, 0.74, 0.74, 0.55, and 1.0 for location of lymphocytic infiltrate, plasma cell prevalence, polyethylene prevalence, polyethylene size, metal prevalence, and deformation of the humeral polyethylene bushing respectively.
Our results show that conformational change in the polyethylene bushing is strongly associated with large particulate polyethylene debris and a high prevalence of metal and polyethylene particles, confirming that conformational change in the bushing is due to loss of material, not just deformation. Of note, no difference was found between RA and non-RA patients in terms of the severity of device deformation, the composition or magnitude of wear debris, or the average pre-operative range of motion, suggesting that differences in the cellular and tissue reaction cannot be attributed to these factors. The only positive association with less conformational change and wear debris was having a primary TEA revised <4 years after implantation, regardless of underlying diagnosis.
We found an increase in the intensity of lymphocytic aggregates among untreated RA patients, which correlated with the presence of metal and polyethylene particles. This suggests activation of the adaptive immune system in these patients. Analysis of the pattern of prosthetic wear revealed that the ulnar and humeral polyethylene bushings exhibited extensive wear and fragmentation, including metal particle formation related to unintended metal-on-metal wear. Previous studies reported the presence of an inflammatory cell infiltrate and multinucleated giant cells with regions of fibrosis and focal necrosis associated with extra- and intracellular metallic and polymeric particles within histiocytes [3,4]. Only a minimal lymphocytic reaction was found in most cases, and no difference in cellular characteristics was reported between patients with or without inflammatory arthritis. Our study suggests that while no increase occurred in the presence of lymphoplasmacytic infiltrates in patients with inflammatory arthritis, the degree of plasma cell infiltrates was significantly increased among those with diffuse lymphocytic aggregates.
The presence of plasma cells associated with metal wear debris is a characteristic histologic feature in patients undergoing revision surgery after metal-on-metal hip resurfacing [8,9]. Diffuse aggregates of lymphocytes and plasma cells are associated with regions of tissue necrosis and extensive fibrin deposition. The reaction is thought to represent a hypersensitivity to the metal wear products. These findings differ substantially from those in our series and other TEA series [3,4], and may represent a distinct clinical entity in a weight bearing joint. However, most of the patients on anti-TNF therapy in our study demonstrated a similar, high prevalence of plasma cells only in tissue sites with high metal debris, suggesting that metal debris may initiate the plasmacytic response. The presence of lymphocytic aggregates, however, was significantly associated with lower metal particle prevalence. This suggests that while metal might specifically bring about a plasmacytic response, polyethylene may be driving the broader lymphocytic response in these patients. Another possibility is that metal particles in the nanometer range, which have been previously described particularly in metal-on-metal implants and are not counted through visual identification, could cause an underestimation of the true metal particle burden and its contribution to a lymphocytic response .
Fujishiro et al. recently reviewed a large series of patients undergoing revision surgery after total hip replacement for non-metal-on-metal implants . Their series was restricted to patients without known inflammatory arthritis, and included patients undergoing revision surgery for septic joints. They noted lymphocytic infiltrates in half of the patients with aseptic loosening, 62% of whom demonstrated a diffuse pattern of lymphocytic infiltration. They noted that the diffuse pattern was most commonly associated with regions of metal wear accumulation. Only 7% of aseptic cases had plasma cells associated with the lymphocytic aggregates. Their detection of plasma cells in a low number of non-RA patients is similar to our findings, and contrasts with the high prevalence of these cells in our RA patients.
Our findings confirm our earlier observations in RA patients undergoing revision surgery for failed total knee replacement . Although we detected the presence of lymphocytic infiltrates in the tissue from both RA and non-RA patients, the pattern of tissue distribution and extent of the infiltrates differed substantially between the two patient subsets. The infiltrates in non-RA patients were focal with predominantly perivascular localization and few plasma cells. In contrast, in RA patients the infiltrates exhibited both a perivascular and interstitial pattern of distribution and were often organized into sheet-like infiltrates with a high proportion of plasma cells. High polyethylene particle prevalence and large polyethylene particle size was associated with a high plasma cell prevalence within lymphocytic aggregates in RA patients. In contrast, no non-RA patients with many or large polyethylene particles in the peri-implant tissue had lymphocytic aggregates with a high plasma cell prevalence. The prevalence of plasma cells in treated RA patients fell between these two groups, suggesting that the reduction in systemic and tissue inflammation associated with anti-TNF therapy results in attenuation of the pattern of cellular responses observed in RA patients.
Our results provide evidence that patients with RA exhibit a differential cellular reaction to inorganic metallic and polyethylene debris compared to patients without RA. The tissue reaction in the non-RA patients exhibited features of a non-immune granuloma. Lymphocytic infiltrates were scattered within the tissue and limited primarily to a perivascular location. In the untreated RA patients, the presence of more generalized lymphocytic infiltrates containing abundant plasma cells are indicative of participation of cellular components of both the innate and adaptive immune system, a feature characteristic of an “immune granuloma.” These findings recapitulate the observations of Caplan and coworkers who described the presence of atypical pulmonary nodules in a series of coal miners . Histopathologic analysis of those nodules revealed the presence of immune granulomas with extensive lymphocytic infiltration with T and B cells. Importantly, these lesions were absent in miners without RA or in RA patients without exposure to coal dust, leading to the hypothesis that these findings reflected an aberrant immune response to the inorganic components of the coal dust in the RA patients.
The nature of the underlying immunologic mechanisms by which inorganic particles initiate an inflammatory reaction with cellular features of an immune granulomatous response remains speculative. Importantly, this reaction in both peri-implant tissues and “Caplan’s nodules” develops in the absence of articular cartilage. Cartilage matrix components have been implicated as potential immunogens in the pathogenesis of RA synovitis; our observations provide evidence that cartilage is not playing a role in the immune response . The association of anti-TNF therapy with diminished cellular features of the tissue reaction parallels its known effects in reducing inflammatory cell infiltrates in RA synovium .
In addition, there was no difference in the presence of radiolucencies between patients treated and not treated with etanercept. While limited by a small sample size and the unavailability of three-dimensional imaging, these findings are consistent with other studies, which have not found anti-TNF medications to be protective against osteolysis .
Our data show a trend towards the TEA lasting longer in RA patients compared with non-RA patients. This could be due to lower demands placed on the joint replacement by RA patients. However, treated RA patients show a length of implantation similar to non-RA patients and shorter than untreated RA patients. TNF users may have had better-controlled disease, enabling them to resume normal function with higher forces across the joint, accelerating prosthesis loosening.
Our study has limitations, including the small sample size and retrospective data collection. Data were collected cross-sectionally from the time of revision surgery, making it impossible to assign causation. Histopathologic analysis was also limited by the number of available slides. Etanercept was FDA approved in 1998. Patients with more severe RA operated on in the latter part of our cohort may have been more likely to receive a TNF inhibitor, a potential systematic bias. However, the fact that anti-TNF’s were not readily available until early 1999 prevented any confounding by indication for the first three years of our cohort.
In summary, we provide evidence that patients with RA exhibit a differential cellular response to prosthetic wear debris compared with non-RA patients, and that this reaction is mitigated by the use of anti-TNF inhibitors. The immune tissue reaction is similar to that observed in RA patients exposed to silica, another inorganic particle, and specifically consistent with earlier findings of Goldring et. al . Of note, this reaction develops in the absence of articular cartilage, which has been implicated in the immunopathogenesis of RA synovitis [13, 16]. These results provide evidence of a pathogenic activation of the adaptive immune response in RA and could have implications for the treatment of peri-implant osteolysis.
The authors thank Dr. Stephen Lyman for his assistance with the kappa statistic analysis.
Anant Vasudevan was supported by a Yale University School of Medicine Medical Student Research Fellowship, Summer Student Fellowship from the New York Chapter of the Arthritis Foundation, and ACR REF/Abbott Medical and Graduate Student Achievement Award. Dr. Steven Goldring was supported by a grant from the American College of Rheumatology Research and Education Fund, Within Our Reach. Dr. Mandl was supported by a K23AR050607 grant from NIH/NIAMS and CERT Grant from AHRQ U18 HS016075.
Financial Disclosures/Conflict of Interest:
Dr. Steven Goldring has the following commercial financial disclosures: research grants from Boehringer Ingelheim and consulting for Merck Serono, Novartis, Pfizer Pharmaceuticals, Bone Therapeutics, and Roche. Dr. Timothy Wright has the following commercial financial disclosures: research grants from Stryker and Synthes Spine, stock ownership in Exactech, royalties from Mathys ABG, and an editorial honorarium from the Orthopaedic Research Society. Dr. Figgie has one commercial financial disclosure: research grants from Ethicon.