Posterior lumbar interbody fusion (PLIF) is a standard procedure for the treatment of different lumbar degenerative disorders. The goals of fusion procedures using cages/spacers with posterior instrumentation and grafts are to stabilize the motion segment and to facilitate the fusion process. Spacers were initially manufactured using either stainless steel or titanium alloys [1-3]. Metal cages showed good fusion rates [4,5]; however, three problems have been associated with their use. These are subsidence of the cage into the adjacent vertebrae, difficulties in assessing fusion by imaging, and stiffness of the material that reduces the amount of mechanical stimulation to the bone grafts, which might delay the fusion or shielding of the bone graft [6-11]. To overcome these shortcomings, polyetheretherketone (PEEK) cages have been introduced. PEEK cages have several advantages over metal, including the lack of metal allergies and moderate stiffness, which has a modulus of elasticity similar to that of human cortical bone that may increase the union rate of the grafted bone by minimizing stress-shielding effects. Moreover, the radiolucency of PEEK cages results in more reliable evaluations of fusion on computed tomography (CT) images, allowing for a clearer, unobstructed view of new bone formation during the follow-up examination and low numbers of artifacts on magnetic resonance imaging [12,13]. However, there is insufficient evidence supporting the superiority of PEEK cages over metal cages as optimal interbody spacers [14,15]. Only a few articles have recently demonstrated unfavorable radiographic endplate findings with the use of PEEK cages, which may lead to nonunion [15,16].

The aim of the current study is to assess the endplate changes resulting from the use of PEEK cages in PLIFs.

A total of 65 consecutive patients (21 males and 44 females) were enrolled in the study. The average age at surgery was 35.44±19.60 years (range, 26–67 years). The number of levels fixed was 86. The most frequently fixed levels were L4–L5 in 38 levels (44.19%) and L5–S1 in 36 levels (41.86%). The mean follow-up was 16.5±10.1 months (range, 6–57 months). In 41 cases (63.08%), the follow-up was ≥12 months. The patients’ demographic data are summarized in Table 1.

On the basis of our initial observations and recent articles, a high prevalence of unfavorable endplate changes after PEEK application in lumbar fusions surgery exists. We feel that the use of the term “cavity” is more appropriate because of the lack of any histological evidence for the other terms previously used to describe that phenomenon (osteolysis or cysts) [15-17]. Endplate cavities were seen in almost half of the levels in this study (48.84% by observer 1 versus 54.65% by observer 2). This figure was higher than previously reported in studies by Tanida et al. [16] (17.2%) but slightly less than that reported by Nemoto et al. [15] (60%). Despite the significant endplate changes observed in our study, only fair agreement existed between the observers (κ=0.373). Moreover, our results revealed moderate reliability for endplate location (κ=0.414), fair for cavity number (κ=0.281), and poor for cavity size (κ=0.186). We were unable to explain the reason for the differences. However, it should be noted that, on the basis of our results, the definition and characteristics of endplate changes following the use of PEEK should be reevaluated and standardized for better reliability.

Although the mechanism(s) responsible for the formation of vertebral endplate cavities remains unknown, several mechanisms based on biomechanical as well as biological PEEK incompatibility have been proposed. One of them is based on the lower stiffness of PEEK compared to titanium and the lack of sharp teeth on the cage, resulting in insufficient initial stability, which induces micromotions between the cage and the vertebral endplate. Micromotions may lead to microfractures, edema, focal bone resorption, and subsequent local necrotic regions, which may be similar to the formation of subchondral bone cysts in arthritic joints [16]. This explanation seems logical, and it indicated that cavities should be surrounded by sclerosis. Therefore, we evaluated sclerosis in our radiographs. However, our results showed less than poor reliability in the detection of sclerosis. Moreover, the endplate changes might have been vertebral osteolytic changes resulting from a reaction to the PEEK polymer, similar to that induced by cement in arthroplasties [15]. However, we have no pathological-based literature to support that hypothesis. Pre-existing cysts, seen as Modic changes similar to those in arthritic joints, could also be a possibility [16]. In addition, excessive endplate preparation might also be a factor, especially if the procedures were performed by less-experienced surgeons. Therefore, surgical skill is of the utmost importance in performing interbody fusions. Our surgeries were performed by two experienced consulting surgeons. We could postulate two other mechanisms that, to our knowledge, have never been described. The first might have occurred because of the habit of some surgeons to use the largest possible cage, which may cause injury to one or both of the endplates. The second mechanism could be subclinical low-grade infections. Our study excluded patients with frank postoperative infections. However, both mechanisms have to be considered in future studies.

The results of union/nonunion detections in this study should be interpreted with caution because of the relatively short follow-up and the moderate agreement between the assessments. The definite union rate was lower in patients with <24 months of follow-up and higher for those with >24 months follow-up (77.7% and 72.2% for observer 1 and observer 2, respectively).

The previous high union rate (100%) at 24 months reported by Brantigan et al. [18] was based only on the assessment of plain radiographs. Schimmel et al. [19] reported a high number of patients who underwent revision surgeries (n=26, 27.4%) mainly because of symptomatic pseudarthrosis for ALIF using PEEK cages. Nemoto et al. [15] reported a 96% union rate in the titanium group and 64% in the PEEK group at 12 months, which increased to 100% in the titanium group and 76% in the PEEK group at 24 months. They reported vertebral osteolysis in 60% of the cases with nonunion. By contrast, this abnormal finding was not present in the titanium group [15].

In addition, Tanida et al. [16] reported postoperative bone union rates of 75.2% and 74.5% at 1 year and 82.8% and 80.4% at 2 years for the titanium and PEEK groups, respectively. The rates of positive cyst signs in the titanium and PEEK groups were 17.2% and 13.7%, respectively. The nonunion rate with positive cyst signs was 100% in both groups at 1 year [16]. Similarly, we found the percentage of cavities to be higher in patients with nonunions (p<0.0001 for the first observer and p=0.05 for the second observer). We agree with Fujibayashi et al. [17], who reported that the clinical relevance of a vertebral endplate cyst might be an early predictor of nonunion.

Our finding of only moderate agreement in the detection of bony unions might reflect the high dependence and considerable variability between observers despite the criteria applied. In our opinion, it could also mean that the radiological evidence of healing after the use of PEEK cages was not clear or strong enough to reach a consensus. Further CT-based studies could support that hypothesis. Moreover, we do not support the proposed simple evaluation of radiologic fusion in PEEK cages by using only anteroposterior and lateral dynamic radiographs, and those evaluations should be interpreted with great caution.

A clear disadvantage of PEEK cages is their smooth and hydrophobic surface, which results in poorer primary stability. Furthermore, the resulting low surface energy negatively effects bone growth onto the implant. That is why many surgeons use expensive growth factors, such as rhBMP, to improve interbody fusion [20,21].

One strategy to improve primary stability and osteointegration is to modify the implant’s surface by chemical or physical treatment or by adding an osteoconductive layer, usually hydroxyapatite or titanium, thereby increasing the contact area and roughness [22]. Recently, the first author of this study published the first results of a new vacuum plasma sprayed titanium-coated carbon/PEEK composite cage for lumbar interbody fusion. A significant degree of clinical improvement was achieved, in addition to radiological unions in all patients with a mean 24-month follow-up. However, the fusion was evaluated only by plain radiographs [23].

Our study had several limitations. First was the sample size, which needs to be expanded to provide more power and reliability. Second was the retrospective nature of the study. A prospective study could identify the evolution of cavities with respect to fusions. Such a study could help in answering whether the endplate cavities found in the initial follow-up period are predictive factors for future nonunions. Third, unfortunately, we did not include preoperative CTs to exclude pre-existing Modic or arthritic endplate changes because they were available. Fourth, we did not check the bone mineral density in all cases. Correlative studies between bone quality and cyst formation should be conducted in the future [17]. Fifth is that the follow-up period did not extend long enough to study the long-term effects, and therefore, longer follow-up studies are needed. Furthermore, we did not include patient characteristics that were likely to affect outcome and fusion, such as smoking and body mass index. Although this information is important when considering clinical outcomes, this was not the focus of the current study. Finally, two surgeons performed all surgeries in the current study. Although the procedures were consistent in principle, it was beyond the scope of this article to include any further clinical or postoperative implications or correlations.

There was a high rate of unfavorable radiographic findings in the form of endplate cavities in PLIF cases using PEEK cages. A more comprehensive classification for the assessment of fusions and endplate cavities should be formulated. We strongly recommend further CT-based studies with larger sample size and longer follow-up.