Proposal for a classification system of radiographic bone changes after cervical disc replacement

Cervical total disc replacement (cTDR) is becoming an established alternative to fusion for treatment of degenerative disc disease and associated radiculopathy and myelopathy [1‐4]. Starting in the 2000s, the clinical effectiveness of cTDR has been supported by randomized clinical trials, now with intermediate- and long-term follow-up [2, 4‐10]. Early cTDR designs were often based on traditional orthopaedic biomaterials, including polyethylene, CoCr alloys and Ti alloys [3, 10], but the lower biomechanical loading demands of the cervical spine relative to the lumbar spine have also encouraged innovative designers to investigate new bearing materials [11], incorporating metallic alloys, ceramics, polycarbonate urethane, and/or PEEK, with no previous clinical precedent as bearing materials in large total joint replacements. As cTDR designs and its biomaterials continue to evolve, and as utilization of the procedure increases with longer historical exposure, there has been increased attention on identifying mechanisms of cTDR failure [12], such as subsidence, migration, and/or wear, along with recommended treatment paradigms for each clinical failure scenario [12]. Ideally, classification and treatment pathways should be generalized with the understanding that cTDR failures, and their revision approaches, may be design specific.

Since the beginning of large-joint arthroplasty, clinicians and orthopedic researchers have used radiographs to evaluate the status of implant fixation and the likelihood of impending revision surgery. For example, Delee and Charnley [13] and Gruen [14] proposed acetabular and femoral zones in the 1970s which later came to be widely used in describing the location and progression of radiolucent lines and osteolytic lesions around a total hip replacement. Progressive radiolucent lines around a hip component of greater than 2 mm have been found to be associated with clinically relevant loosening that may require revision [13, 15]. However, an equivalent, generally accepted radiographic classification and treatment approach is not yet available in cTDR, in part because of the recent adoption of this relatively new procedure, and also because of the diversity in implant designs. The clinical situation is more complex in cTDR than in large joints, in which diagnosis and treatment are generally based on radiographic examination as the “keystone” diagnostic tool [15]. By contrast, cTDR treatment decisions involving revision are less frequently undertaken based on radiographs alone, without further workup that may include computed tomography (CT) scans and/or magnetic resonance imaging (MRI). Of these imaging modalities, CT is the most reliable test for bone loss.

Nevertheless, plain radiographic assessment remains the front-line patient assessment tool for cTDR [12, 16]. Previous studies have underscored the need for classification of radiographic changes around cTDR including heterotopic ossification (HO) [17, 18] and bone loss [12, 19]. Bone loss around orthopedic implants can result from sepsis [20, 21], or aseptically due to bone adaptation or remodeling due to stress shielding (sometimes referred to as “Wolff’s law”) [22], osteoporosis [23], fluid pressure [24], as well as the well-described chronic inflammatory reaction to wear debris termed “osteolysis” in peripheral joint arthroplasty [19]. Today, the etiology of bone loss around large-joint implants can be effectively elucidated with plain radiographs and long-term clinical experience with specific designs [15], but whether a radiographic bone lesion is septic or aseptic, or due to bone adaptation or particulate wear debris, may require confirmation by pathologic examination of retrieved periprosthetic tissues [25] as well as microbiological cultures of synovial fluid and periprosthetic tissues.

However, authors in the spine field often use terms such as wear, osteolysis, and bone loss interchangeably to describe radiographic changes around cTDRs, without pathologic confirmation for the etiology of bone loss [19, 26, 27]. Zavras et al. [12] have recently proposed a cTDR failure classification system, with wear as a type of failure with subclassification of osteolysis with minimal or severe bone loss. The ambiguity and lack of specificity in the radiographic assessment of periprosthetic bony changes around cTDR is a barrier to effective scientific communication and the development of effective treatment recommendations.

Consequently, the purpose of the present study was to develop and assess a radiographic classification system for reactive changes after cTDR. We sought to answer the following principal research question: can plain radiographs be used for accurate and repeatable classification of bony changes after cTDR? To answer this question, our team developed and evaluated a classification system using blinded and de-identified radiographic case studies drawn from their clinical experience with multiple implant designs.

The overall agreement of assessments ranged from 49.9% (45.1–73.1%) to 94.7% (86.9–100.0%) (Table 1). There was reasonable agreement on the presence or absence of bone loss or radiolucencies (range: 58.4 (51.5–82.7%) to 94.7% (86.9–100.0%), Table 1), as well as in the progression of radiolucent lines (82.9% (74.4–96.5%)). However, the lowest agreement was found on the assessment of severity of bone loss (49.9% (45.1–73.1%)) or the progression of bone loss (66.2% (59.7–87.7%)). High overall agreement was achieved in binary assessments of HO (range: 75.2 (66.9–91.5%) to 91.7% (83.2–100.0%)) and implant function, such as the presence of subsidence (82.2% (73.1–94.9%)), migration (90.3% (83.2–100.0%)), loss of core implant height (90.9% (83.2–100.0%)), and olisthesis (72.0% (64.8–91.7%)).

As the usage of cervical arthroplasty has increased over the past 20 years, the prevalence of clinical and device-related failures associated with this technology has also increased, motivating the present study to develop a clinically useful classification system to aide clinicians with assessing bony changes over time. In the present study, we developed a robust classification scheme for bony changes around cTDR that was evaluated across a broad range of contemporary implant designs. We utilized serial radiographs of routine clinical cases in our radiographic assessment of cTDR. It became increasingly clear during our collaboration how important serial radiographs are in accurately assessing any progression of bone loss. We routinely found evidence of bone-implant gaps upon assessment of post-operative radiographs that did not fill in or remodel up to five years of implantation. Without access to serial radiographs, those initial unresolved bone-implant gaps could be misinterpreted as bone loss. Therefore, if only a single radiographic observation is available, clinical interpretation is problematic, and additional imaging studies, including computed tomography (CT) and magnetic resonance imaging (MRI) are strongly recommended prior to undertaking further clinical intervention.

There are relatively few studies that have addressed classification of cTDR failure modes that focus specifically on the bone-implant interface [12]. Zavras et al. [12] proposed a general classification for cTDR reasons for revision, that specifically focused on failures that required revision surgery. Zavras and colleagues also classified revision due to septic loosening as a sub classification of infection, whereas revisions due to implant wear were subclassified as either with or without osteolysis. The implication of Zavras’ classification system is that osteolysis is solely associated with wear-related failure of cTDR. We certainly agree with Zavras’ identification of wear as a potential mechanism for osteolysis in the spine, however given that other mechanisms can result in periprosthetic bone loss, such as infection and stress-shielding, we underscore the importance of classifying bone loss based on imaging without attribution of etiology until histopathological confirmation of the root cause has been determined [19]. In large joint orthopaedics, decades of early fixation methods and historical polyethylene bearing materials resulted in radiographic classification systems for bone loss that reflected a time when particulate wear debris from bone cement, metal, and polyethylene was responsible for many more clinical failures than infection. A similar situation does not occur today with cTDR, in which wear related revisions and infection as documented in prospective randomized clinical trials [2, 5‐9], are both equally low in incidence, and hence the etiology of periprosthetic bone loss associated with cTDR cannot be reliably assumed based on radiographs alone. In hip and knee replacements, clinical experience has evolved over decades to educate the intuition of surgeons with the radiographic interpretation of bone loss due to non-inflammatory bone adaptation in comparison with inflammatory mechanisms of osteolysis such as infection and particulate wear debris [19]. However, this is not the case with cTDR in which the clinical experience with radiographic interpretation of periprosthetic bony changes is still relatively early in its development. Although osteolysis around cTDR has been raised as a potential clinical concern [19, 27], the use of a reliable radiographic classification system as proposed here, coupled with device and tissue retrieval analysis, should improve future scientific communications about bone loss around total disc replacements, along with their reasons for revision.

Radiographic outcomes are publicly reported for cTDR designs that have completed clinical trials as part of their regulatory approval process by the United States Food and Drug Administration (FDA) (https://​www.​accessdata.​fda.​gov/​scripts/​cdrh/​cfdocs/​cfpma/​pma.​cfm). Radiographic outcomes and observations were extracted from the summaries of safety and effectiveness (SSED) from 12 FDA-approved cTDRs currently on the US market and summarized in Table 2. Overall, these findings included assessments of heterotopic ossification, radiolucencies, disc height, migration, subsidence, and loosening (Table 2). However, the reports lacked assessments of bone loss, and a range of methodologies was used by sponsors in radiographic assessments of cTDR, highlighting the need for standardization in this area. Similar findings have been reported for longer-term radiographic studies: a recent systematic review and meta-analysis aggregated all available 5 + year clinical outcomes for cTDR and reported that observations of bone loss/osteolysis were largely absent (or perhaps reported under other terminology or conditions) [32]. Only one reviewed study identified osteolysis, which was reported for six patients out of 715 [33]. Similarly, a recent analysis of shorter-term results found a low number of studies reporting osteolysis [34], and a review regarding cTDR radiological outcomes including HO lacked guidance on osteolysis/bone loss appearance [35]. While reports of radiolucencies were more common in the SSEDs and clinical trials, explanation of the assessment criteria used were often lacking. In our assessment, we considered progression of radiolucencies to be an additional and more telling indicator of the bone-implant interface. Thus, our study adds to the body of clinical findings by highlighting a bone loss metric that was perhaps poorly understood when these trials were designed and warrants continued research even today.

Using the proposed classification system, we observed reasonable agreement among experienced investigators. Previous researchers have employed classification systems for anterior bone loss around cTDRs [28, 29], rather than assessing the anterior, centralized, and posterior regions of the endplates as we proposed in the present study. In addition, previous investigators did not assess the repeatability or reproducibility of their classification systems for anterior bone loss, making it difficult to compare prior research with the present study. Chen and colleagues [29] evaluated radiographs of the Bryan disc for anterior bone loss, with a grade ranging from 0, with no bone loss, to 2 for “obvious bone regression.” They found that anterior bone loss with a Grade of 1 or 2 occurred in the inferior endplate of 54/121 patients (44.6%) from their series within the first 6–12 months after surgery [29]. Kaiser et al. [28] performed radiographic measurements of the relative distance of anterior bony coverage by the endplates of three different cTDR designs over time. The researchers assessed the anterior bone loss as minor (< 5%), moderate (5–10%), and severe (> 10%). Among the 156 cTDR evaluated, Kaiser [28] found anterior bone loss in 57.1% of cases evaluated. In our study, we found that binary radiographic assessments, such as the presence of absence of bone loss in a particular location, as well as subsidence, migration, and progressive core height loss, were found to exhibit the highest agreement among investigators. Lower agreement was found in our study among investigators when grading the severity of bone loss, indicating the lack of consensus as to what constitutes mild, moderate, or severe bone loss. For example, Kaiser [28] considered 10% of anterior bone loss to be severe, whereas the investigators in the present study considered loss of 1/3 of the endplate coverage to be severe. With regard to the accuracy of the proposed classification system, we expect that broad use of the tool by experts will help establish “gold standard” classifications which can be used to assess validity in future work. Our study confirms the utility of the HO classification system proposed by McAfee et al. [18] Because HO can occur in some regions of a cTDR that also are undergoing bone loss, we consider it important to include HO assessment in the characterization of bony changes using radiographs.

We would like to underscore several limitations for the reader. In the present study, we focused on developing a classification system for plain radiographs because of their ubiquity in clinical follow up of spine patients. However, plain radiographs can underestimate bone loss. Variations in imaging technique, such as tilting or obliquity of the cTDR with respect to the plane of radiographic study, greatly complicate the assessment of the bone-implant interface. Differences in radiographic exposure over time can also confound the interpretation of the bone-implant interface, if insufficient or inconsistent beam penetration of the bone occurs over time. Additionally, the presence of bony or radiographic abnormalities can also affect the interpretation of the radiographs. These limitations may help explain, in part, the lower agreement in the assessment of bone loss severity among investigators, and lower-quality radiographs may complicate use of the classification system in practice. In addition, plain radiographs have limitations in diagnosis of the underlying causes of the bone loss, and whether the mechanism in a particular patient is due to a septic or aseptic etiology [25]. Furthermore, it can be challenging to assess whether loosening is caused by failure to achieve initial stable fixation of the metallic endplates. Although the progression of radiolucencies can be useful in this regard, subtle bone-implant relative motion may also be appreciated by alternating flexion and extension views while the images are stabilized on the superior and/or inferior vertebral body. We did not consider such flexion–extension image assessment as part of the present study. We similarly did not consider AP view radiographs, which restricted our ability to clearly explore changes occurring lateral to the implants. Despite these limitations, due to the prevalence of radiographs in cTDR patient care, improving the consistency in terminology and classification of radiographic changes at the bone-implant interface is expected to improve communication among clinicians.

In summary, a standardized nomenclature for bony changes following cTDR will facilitate accurate and reproducible scientific communications regarding the clinical outcomes of this procedure. The novel system proposed demonstrated good concordance among experienced investigators in this field and represents a useful advancement for improving reporting in cTDR studies.