The lexicon for periprosthetic bone loss versus osteolysis after cervical disc arthroplasty: a systematic review

Two of the authors (J.M.W. and S.N.S.) systematically searched electronic databases following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [24]. Specifically, a comprehensive search of the PubMed, Google Scholar, and ScienceDirect databases was conducted for studies reporting outcome of CDR including evaluation and assessment of periprosthetic bone loss or osteolysis. The keyword search terms used were “cervical disc replacement/arthroplasty,” “osteolysis,” “bone loss,” “radiograph,” and “complications.” To be included, studies must include a radiographic review, incidence of bone changes, and be written in English. Studies were categorized as prospective, randomized, or retrospective reports.

The following categories of data were extracted for further analysis from each article: (1) general information such as type of device, patient age, and any patient reported clinical outcomes (2) data pertaining to specifics of bone loss or osteolysis, such as time of observation, implant materials, subsequent surgery, and explant analysis results, (3) patient-reported clinical outcomes, and (4) incidence of bone loss or osteolysis. Additionally, any specific information pertaining to bone loss grading criteria, such as that of Kieser et al., was included [25]. In prospective studies in which the number of patients lost to follow-up was not provided, 100% follow-up was assumed.

In a number of studies, the terms “bone loss” and “osteolysis” were used interchangeably, exemplifying ambiguity that the present study aims to address and clarify. In these cases, the present study’s authors analyzed each articles descriptions, provided images, and clues to assess whether the outcome being reported could be deemed as bone loss or osteolysis beyond a reasonable doubt.

The searches yielded 1061 articles. Abstracts were screened and reviewed and, after removing studies that did not qualify or were duplicates, 49 articles were further evaluated (Fig. 2). Exclusion criteria included studies lacking bone loss or osteolysis data, focusing on the lumbar spine or anterior cervical discectomy and fusion data only, or lack of radiographic data. Inclusion criteria included fell-text studies assessing patients with a CDR, radiographic evidence of bone loss or osteolysis, and cervical spine data. Nine clinical articles were further excluded, leaving 40 full-text articles to be reviewed. Another 17 articles did not assess bone changes radiographically and were excluded, leaving 23 articles for the present review (Table 1).

From the 23 included articles, the following information was collected: author(s), number of patients, follow-up, study type, mean patient age, country of origin, and overall reoperation rates (Table 1). Studies that included multiple devices were then separated according to implant, by implant name, implant materials, number of patients, follow-up time, and reported term(s) of bone change. The majority of cases reporting osteolysis distinguished between non-inflammatory bone loss and inflammatory osteolysis, whereas articles focused on bone loss were more inconsistent in their terminology and definitions. Overall, nine articles used bone loss and osteolysis interchangeably to describe the respective bone remodeling seen (Tables 2 and 3). Additionally, when reported, progression, removal, and implantation duration were noted. Results for histology or explant analyses were recorded when presented (Tables 4 and 5).

Overall, fourteen articles reported the resorption of periprosthetic bone due to osteolysis in six implant designs (Table 2). Duration of implantation prior to reoperation ranged from 15–96 months. Radiographic analysis was performed at multiple time points, with progressive osteolysis observed in eight articles [26‐32]. In twelve articles, the implants were removed. In one article, a case report for a 2-level disc replacement, a ventral mass was removed, the devices were retained, and a 3-level posterior fusion was performed [33].

Early reoperation rates were reported for the devices composed of stainless steel or CoCr components, previously reported to be associated with metal sensitivity. Specifically, the Prestige-ST, composed of steel plates and steel core, was reported to have high incidence of osteolysis between 7 and 30 months [28, 34]. Osteolysis was noted later for devices with polymeric components, ranging from 24–120 months following implantation (Table 4).

In eight articles, histological analysis was reported to confirm the radiographic and/or intra-operative diagnosis of osteolysis [29‐33, 35, 36]. Two articles reported positive cultures for Propionibacterium acnes[29, 30] and two for Staphylococcus epidermidis.[35, 36] Some of the observations included abundant giant cells, granulomatous formation, macrophages, metal particles, and flakes of polymeric debris, which are features of a potentially osteolysis-inducing foreign body response [23].

Neck pain was most frequently reported for patients with osteolysis [26, 27, 30, 33‐38]. Radicular symptoms were reported in four studies[28, 33, 35, 38] and paresthesia was reported in three case reports [29, 30, 33]. Other symptoms accompanying osteolysis were: dysphagia, respiratory compromise, myelopathy, shoulder pain, distal upper extremity numbness, and decreased fine motor dexterity. Clinical outcome scores were reported in two case studies with VAS and NDI scores both reporting severe pain [36, 38].

Nine articles reported non-inflammatory bone loss in seven implant designs (Table 3). In the majority of studies, the location of bone loss was specified as the anterior region of the vertebral bodies [39‐42], occurring 3–6 months post-operative. Only one article reported failure of an implant with severe bone loss; however, other adverse events were reported in that case, so it was unclear what the main cause of revision was [43]. In contrast, the remaining eight articles reported no progression or correlation of bone loss and device failure (Table 5). For these studies, mean length of follow-up ranged from 24–120 months; therefore, some studies may not have sufficient follow-up time to assess long-term ramifications of early bone loss.

Five articles reported range of motion (ROM) data for the patient cohorts [39‐42, 44]. Of these, only one study compared ROM of patients with or without bone loss, reporting mean pre- operative ROM of 8.99° and final follow-up of 8.25° in patients with bone loss, and pre-operative ROM of 9.1° and post-operative of 7.09° in those with no-bone loss, but this difference was not statistically significant [44]. The remaining articles reported ROM for the total patient population, but did not distinguish between patients with or without bone loss [39‐42]. Combining the results of these nine articles, the ROM at final follow-up was comparable to the pre-operative ROM (P = 0.2).

Among included articles, 515 of 949 (54.2%) patients experienced bone loss of some severity. One article was excluded from this calculation due to duplicity of the patient cohort from another included article. Five articles used grading criteria to rank bone loss, with mild bone loss most commonly reported [25, 39‐41, 44]. Bone loss differed in both severity and incidence among devices, but because data was not reported consistently, no definitive conclusions could be drawn.

Clinical outcome scores for NDI, VAS arm and neck were measured for patients with bone loss in three articles [39, 40, 44]. However, a combined average could not be calculated due to differences in reporting methodologies. For example, one study reported VAS arm and neck scores using numerical values while another reported scores qualitatively, in terms of minor and major changes [39, 40]. Nevertheless, each concluded that bone loss did not have an effect on clinical outcome scores.

Pain symptoms were reported in four articles; however, three improved over the course of follow-up [39, 41, 43, 45]. Two studies reported early neck pain with long-term improvement [39, 45]. One study reported radicular pain with long-term resolution [41]. The final study reported recurrent neck and arm pain with no resolution until revision. This was the only study reporting bone loss in a failed device [43].

The present study emphasizes the stark differences between osteolysis and non-inflammatory bone loss. Inflammatory osteolysis may present around the implant and appear as large lesions being scooped out of the bone (Fig. 1a&1b). By contrast, non-inflammatory bone loss appears as an erosive remodeling of the vertebral body (Fig. 1c&1d). Osteolysis may be present surrounding the implant, whereas bone loss is more common around the peripheral or anterior side of the vertebral body. These clear visual distinctions demand new terminology that accurately describes these differences.

In larger, synovial joints, wear debris has long been recognized as a potential cause of osteolysis. However, in several studies in the present review, the early onset of bone resorption led authors to conclude that wear debris was an unlikely primary cause [39, 41, 44, 45, 53]. In late onset cases, progressive osteolysis was attributed to mechanical failure of the M6-C in five studies, despite findings consistent with device failure and histopathology confirming the presence of particulate polymeric debris and giant cells [29, 30, 33, 35, 36]. This device was noted to include a sheath intended to contain and decrease debris migration, and rupture of the sheath contributed to particle-induced osteolysis [35, 36]. These findings were consistent with those reported for another annulus-nucleus analog device with a sheath, the Bryan [54].

The use of histopathology in confirming the presence of wear debris is pivotal in understanding the etiology behind osteolysis. As with most joint arthroplasties, material selection was crucial for minimizing debris and failures. Although cervical disc replacements utilize similar materials, little is known about how the body responds to these materials in the spine. Unlike most large joints, the spine is not a synovial joint so we cannot use these examples as a precedent. This suggests that material reactions may differ slightly from those observed in other joints and may affect the clinical outcome. Therefore, cases with suspected osteolysis should prioritize histopathology tests to confirm the presence or lack of wear debris in the surrounding tissues.

In the articles reviewed, stress shielding was frequently cited as a cause of non-inflammatory bone loss. Referencing Wolff’s Law, which states that bones will adapt to applied stresses [55], some articles hypothesized that the implants created higher stresses posteriorly, resulting in remodeling to adapt to the modified stresses [45, 53, 56]. Some articles suggest that resection of the anterior longitudinal ligament (ALL) contributes to reduced stresses in the anterior vertebra after surgery [25]. The common “Smith-Robinson” surgical approach for CDR includes this resection which may increase lordosis, decreasing stresses anteriorly [57]. The ALL increases stability and provides a natural constraint for motion [58]. Radiographic observations of anterior bone loss could result from a change in natural motion and stress distribution after ALL resection. Wang et al. reported that patients with bone loss experienced more lordosis following surgery, possibly due to the resection [42]. They hypothesized that this change in stress patterns left a hypo-pressure area on the anterior portion of the vertebra, leading to resorption. Similarly, Kieser et al.[25, 40] reported the resection of the ALL caused bone loss due to decreased anterior stresses. Although these authors believe the timing is inconsistent with this theory, spinal vertebrae are predominantly composed of trabecular bone, which may remodel much faster than cortical bone surrounding hips or knees.

As most bone loss cases have not been associated with pain symptoms or device failure, this etiology would be consistent with remodeling in the cervical spine in the short post-operative time period while the bone adapts to change in stress and is non-progressive. In order to confirm this, longer follow-up with bone loss patients, as well as biomechanical studies to analyze the force distribution of a healthy and implanted cervical spine, are necessary.

Micromotion is another mechanism that has been attributed to bone loss[49], largely resulting in bone loss immediately anterior or posterior [44]. One study suggested that early micromotion during the first 3–6 months lead to bone loss, but after 6 months, no further change was seen, presumably due to a reduction in micromotion through enhanced fixation [41]. Another study reported that patients with hybrid surgeries who were immobile during the first three months had reduced micromotion and a lower incidence of bone loss [44]. The authors suggested that bone loss may have occurred during the process of osseointegration but stopped following successful integration. It has been widely accepted that cyclic motions under 40–150 microns will allow for bone formation [59]. However, as there are no studies that have quantified micromotion to assess the early fixation of cervical disc replacements, it is unknown if the same threshold can be applied to spine devices.

Several authors speculated that infection was the cause for observed osteolysis; yet the majority presented incomplete evidence to confirm infection. Most articles did not provide laboratory results to confirm the infection diagnosis[60] and those that did had conflicting results. Cultures tested positive for either for Propionibacterium acnes or Staphylococcus epidermidis; however, there were reports of some negative cultures or normal inflammatory markers in these cases. Due to this inconsistency, most articles could not conclude that infection was the primary cause [26, 29, 36].

Propionibacterium acnes is well documented to be of low virulence and slow growing, which renders it difficult to culture and diagnose as problematic. Its slow growth requires lengthy periods of incubation, which can lead to contamination and false negatives, further adding to the confusion [61]. Propionibacterium acnes has been reported to show poor sensitivity to standard lab markers such as white blood cell counts and erythrocyte sedimentation rates, making it difficult to detect [62]. Additionally, some positive cultures presented with normal inflammatory markers, suggesting inflammatory markers may not even be sensitive enough in the cervical spine to confirm infection. Therefore, conflicting evidence of an infection associated with osteolysis is expected, making it difficult to confirm an association.

In the absence of a confirmed diagnosis of infection, an immune-mediated response due to metal or polyethylene hypersensitivity was considered a possible cause of the osteolysis [26, 27]. Although hypersensitivity has long been observed in orthopedic implants, no validated test is available to confirm this [63, 64]. Studies that could not confirm infection concluded that wear debris was the main cause of osteolysis, despite no explant images or histological quantification of wear debris being provided. Evidence like this further questions reliability of causation analysis and demands a change.

This review had several limitations, partly due to gaps in the literature for radiographic bone changes. The reviewed articles included many single case reports or small patient cohorts, limiting confidence in the combined averages tabulated. However, nearly all reported cases of osteolysis that could be attributed to reaction from wear debris were in single case reports. This finding alone emphasizes the importance of including all studies in the present systematic review, and not only randomized clinical trials. Specifically, some previous review articles that have addressed this issue have limited their study to randomized trials and may have severely under-reported the incidence and importance of osteolysis.

Studies were not uniform with their grading system or terminology. In some studies, a large number of cases were reported without clearly specifying how bone loss was assessed, quantified, or reported. Future studies would benefit from a universal grading scale of bone changes and radiographic reviews of larger patient cohorts with longer follow-up.