In the treatment of patients with symptomatic degenerative lumbar spine disorders, improving sagittal balance, restoring lumbar lordosis, and decompressing neural elements are critically important. Lumbar interbody fusions achieve these goals by facilitating the placement of large grafts that are capable of increasing the segmental disc angle to restore lordosis and of increasing the disc height for indirect decompression of foraminal and central stenosis. Interbody fusions of the lumbar spine can be performed via anterior, posterior, and lateral approaches, each of which has benefits and drawbacks.¹

Lateral lumbar interbody fusion (LLIF) is a widely used surgical technique that avoids some of the approach-related complications of anterior lumbar interbody fusion² while allowing for larger implants than posterior or transforaminal lumbar interbody fusions.¹ A potential complication of any interbody fusion procedure, including LLIF, is graft subsidence into the adjacent vertebral bodies, which can lead to a loss of disc height, segmental lordosis, and indirect decompression (Fig. 1). Graft subsidence has been shown to be a risk factor for revision surgery.³ However, the factors associated with subsidence after LLIF are incompletely understood.^4,5 The size^6–9 and material^10–13 of the implant, as well as the patient’s bone density,^8,9,14–16 have been implicated in subsidence after interbody fusion.

Examples of interbody subsidence at L3–4 (A) and at L2–3 (B). In both cases, the interbody graft has subsided into the inferior endplate of the grafted level.

We reviewed a case series of consecutive patients who underwent LLIF procedures performed by a single surgeon. We examined the association between radiographic subsidence and age, sex, bone density, cage size and type, and posterior instrumentation. We hypothesized that the prevalence of subsidence may vary with novel titanium (Ti) implant material because of that material’s modulus of elasticity.

Methods

This study was approved by the institutional review board at St. Joseph’s Hospital and Medical Center (Phoenix, AZ). We performed a retrospective review of a prospectively maintained database of consecutive patients who underwent transpsoas LLIF at a single institution performed by a single surgeon between August 1, 2017, and January 31, 2020. Details of the surgical technique have been published elsewhere.^17,18 A total of 131 consecutive patients were identified who underwent surgery at a total of 204 levels. Indications for surgery included spondylolisthesis, adjacent-segment disease, adult degenerative scoliosis, and degenerative disc disease. Patients with infectious and neoplastic indications were excluded. Preoperative standing scoliosis radiographs, lumbar CT, and MRI were obtained for all patients. Postoperatively, standing scoliosis radiographs, lumbar CT, and MRI were obtained for all patients on postoperative day 0–2, and additional standing scoliosis radiographs were obtained at 6 weeks, 12 weeks, 6 months, and 1 year after the operation.

Bone mineral density was estimated on the basis of lumbar trabecular CT attenuation from preoperative CT, as described elsewhere.¹⁹ Additionally, preoperative CT scans were assessed for the presence of vacuum disc and to calculate the area of the inferior endplate at each target level. Implant characteristics (cage width, height, length, and material) were collected from the operative notes and implant logs. Implant surface area was obtained from the manufacturer (NuVasive, Inc.), and the ratio of implant area to inferior endplate area was subsequently calculated by dividing the former by the latter.

Each patient underwent LLIF performed by a single surgeon, as described elsewhere.²⁰ Of a total of 204 operated levels, 137 underwent pedicle screw fixation during the procedure. The use of pedicle screw fixation was determined by surgeon preference but typically reflected concerns for instability or malalignment that could not be corrected with cage placement alone. Implanted cages were either Ti (Modulus; NuVasive) or polyetheretherketone (PEEK) (CoRoent; NuVasive). All cages were packed with bone cellular matrix allograft (Osteocel Plus; NuVasive). Bone morphogenetic protein (rh-BMP2) was not used in any of the cases.

Subsidence was graded according to the Marchi classification.²¹ Standing lateral radiographs were reviewed to determine the extent of cage subsidence into the adjacent endplates. According to the Marchi scale, grade 0 indicated subsidence resulting in 0%–24% loss of disc height; grade I, 25%–49%; grade II, 50%–74%; and grade III, 75%–100%.

For each patient, subsidence was graded on the basis of the most recent imaging. If subsidence was observed, previous imaging studies were subsequently reviewed to determine the timing of subsidence. If it was observed on immediate postoperative images, then the intraoperative imaging was reviewed to determine if subsidence occurred intraoperatively. Cases of subsidence that occurred intraoperatively were subsequently reclassified as intraoperative endplate disruption.

Univariate analyses were used to assess the association between subsidence and age, sex, body mass index (BMI), disc space level treated, cage height, cage material (Ti vs PEEK), ratio of cage area to inferior endplate area, posterior fixation, and CT attenuation of lumbar trabecular bone. A chi-square test was used for categorical data, and a 2-tailed t-test was used for continuous data. After completing the univariate analysis, a multivariate analysis including all variables was performed via a binomial logistic regression.

Results

A total of 131 patients were identified who underwent surgery at a total of 204 levels. Subsidence was initially identified at 23 operated levels (11.3%). Subsidence was graded according to the Marchi classification. We did not observe radiographic subsidence (grade 0) for 181 levels (89%). All 23 levels with subsidence had Marchi grade I radiographic subsidence; we did not observe any grade II or grade III subsidence. On further imaging analysis, 11 (47%) of these levels had intraoperative subsidence, which was identifiable from intraoperative imaging. These levels were subsequently reclassified as having subsidence attributable to intraoperative endplate disruption and were grouped with the 181 levels without subsidence for analysis; thus, the total nonsubsidence group comprised 192 of 204 (94.1%) levels. Twelve of 204 (5.9%) levels with so-called true subsidence were attributed to postoperative cage settling for this analysis. In total, 11 of these 12 levels showed subsidence within 6 weeks after surgery, and 1 of 12 levels showed subsidence between the 6-week and 12-week follow-ups. No new cases of subsidence were observed after 12 weeks.

Levels with true subsidence (n = 12) were compared to those with no subsidence (n = 192) by using a univariate analysis to assess characteristics associated with subsidence (Table 1). Among patients with true subsidence, the mean follow-up was 260 days (range 40–611 days), compared with 226 days (range 40–697 days) in the nonsubsidence group (p = 0.57). The mean age in the subsidence group was 66.2 years, compared with 67.6 years in the nonsubsidence group (p = 0.56). Among female patients, 7 of 114 levels (6.1%) were found to be subsided, compared to 5 of 90 levels (5.6%) in male patients (p > 0.99). The mean BMI was not significantly different between the two groups (31.1 vs 28.5, p = 0.14). Cage height did not differ significantly between the two groups (p = 0.48). PEEK cages were significantly more likely than Ti cages to be associated with subsidence; 12 of 149 (8.1%) PEEK cages were associated with subsidence, compared with 0 of 55 (0%) Ti cages (p = 0.04). The ratio of implant area to inferior endplate area was also found to be significantly lower in the subsidence group (0.34 vs 0.42, p < 0.01). Posterior fixation was associated with a 4.4% (6/135) prevalence of subsidence versus 8.7% (6/69) for those without posterior fixation (p = 0.23). Finally, the mean bone density as estimated by CT attenuation of the lumbar spine was not statistically different between groups (132.2 Hounsfield units [HU] in the nonsubsidence group vs 126.3 HU in the subsidence group; p = 0.63).

A binary logistic regression was performed, which incorporated all variables tested on univariate analysis. The only variable found to have a significant effect on subsidence rates was the ratio of implant area to inferior endplate area (p < 0.01). Increasing ratios were associated with a decreased likelihood of subsidence.

Discussion

We performed a retrospective review of 131 patients who underwent LLIF at 204 levels at a single institution, performed by a single surgeon. The overall prevalence of true subsidence was low at 5.9% of levels treated; subsidence, when observed, was low grade in all cases.²¹ This subsidence prevalence is lower than previously reported prevalences of 10% to 16% among patients with LLIF.^3,4,22,23 Only 1 patient in the subsidence group underwent reoperation; in this case, the patient, with a long-segment construct, presented with a rod fracture 2 levels above the subsidence level. Interestingly, we found that approximately half of the operated levels initially classified as involving subsidence were actually levels in which subsidence was attributable to intraoperative endplate disruption. A review of intraoperative fluoroscopic imaging demonstrated that, in these cases, the endplate was violated with either the Cobb distractor or during placement of the cage itself. This underscores the importance of careful surgical technique as a key step for preventing subsidence during LLIF. In patients with postoperative subsidence, nearly all occurrences were observed within the first 6 weeks after the operation.

In contrast with a previous study, we found no difference in subsidence associated with cage height.⁷ This finding could reflect a difference in surgical technique or in the patient population. At our institution, to decrease the possibility of subsidence, we rarely place implants greater than 10 mm in height. This practice may have limited the ability of our study to detect a difference in subsidence prevalence that would result from cage height. Similarly, we found no difference in subsidence associated with bone quality. Previous studies have demonstrated that bone density is a predictor of graft subsidence;¹⁵ however, our cohort included few patients with osteopenia or osteoporosis, which could bias our findings.

When examining patients with true subsidence, we found on univariate analysis that the prevalence of cage settling was significantly lower among patients with Ti cages (0%) than among patients with PEEK cages (8.1%). However, on multivariate analysis the effect of cage type on subsidence became nonsignificant, and the ratio of the implant area to the area of the inferior endplate became the only significant predictor of subsidence. We excluded the cases of intraoperative endplate violation as a cause of subsidence to isolate the true radiographic outcomes associated with cage settling. Given the findings of the multivariate analysis, the most likely explanation for the different rates of subsidence is that the Ti implants used in this study (Modulus) have a larger surface area than the PEEK cages used (CoRoent), possibly spreading the force over a larger area and decreasing the maximal pressure applied to the adjacent endplates. Another possible explanation is that the beam-hardening artifact from the denser Ti cages may decrease our ability to identify subsidence at the endplate–graft interface. In support of this notion, a small study of Ti cages found a subsidence prevalence of 3.4%,²⁴ which is substantially lower than prevalences reported elsewhere in the literature. Future work may be required to determine a better methodology with which to assess this radiological shortcoming as the use of Ti implants increases in spine surgery.

With respect to concomitant posterior spinal instrumentation, the prevalence of subsidence among patients with versus without posterior fixation was not statistically significantly different on either univariate or multivariate analysis. Biomechanical studies suggest that posterior instrumentation results in stress shielding of interbody grafts and decreased load across the bone–graft interface.²⁵ Theoretically, this could lead to lower graft subsidence.³ Nevertheless, the role of posterior instrumentation in subsidence after LLIF is controversial. One study demonstrated no difference in the prevalence of subsidence,²⁶ although another found a higher prevalence of subsidence associated with the use of stand-alone constructs.²¹ Moreover, a selection bias clearly exists, because the application of and decision to use posterior instrumentation reflects the surgeon’s concern for instability, malalignment, or poor bone quality. Therefore, although pedicle screw fixation may reduce the prevalence of subsidence associated with LLIF, potential confounding may exist.

Our study has limitations. As a retrospective study, this work cannot eliminate potential biases associated with how patients were assigned to the Ti or PEEK group. Because the Ti cage used in this study was only recently approved by the US FDA, patients in the Ti group were typically treated more recently than those in the PEEK group. Our results may be confounded by the refinement of surgical technique over the course of the study.²⁷ This study includes patients with a wide variety of pathologies (degenerative disc disease, spondylolisthesis, and adult degenerative scoliosis) treated with a range of surgical constructs (stand-alone, short-segment, and long-segment constructs). Pooling patients together may obscure factors contributing to subsidence in these varied populations. Because they are based on a single-center, single-surgeon study, our findings may not be representative of different surgeons, institutions, and patient populations. Finally, because this was a radiographic-only study, we did not include any patient-reported outcomes. In our experience, although cases of subsidence that require reoperation are very rare, the variation in patient-reported outcomes across patients with different pathologies is substantial enough to make comparing outcomes challenging. These limitations can be addressed with larger, multicenter, prospective studies.

Conclusions

The prevalence of subsidence associated with modern LLIF surgery remains low. The majority of subsidence cases in our series occurred within the first 3 months after surgery and are classified as grade I. Based on the findings of this study, the prevalence of subsidence after LLIF is lower among patients with Ti cages than among patients with PEEK cages, although this finding is most likely attributable to the relatively larger surface area of the Ti cages used in this series. Additional studies are necessary to further define this relationship.

Acknowledgments

We thank the staff of Neuroscience Publications at Barrow Neurological Institute for assistance with manuscript preparation.

Disclosures

Dr. Uribe reports being a stock shareholder, grant/research support recipient, and consultant for NuVasive, Inc., as well as a consultant for SI-BONE, Inc. and Misonix, Inc.

Author Contributions

Conception and design: Uribe, Ohiorhenuan, Walker, Zhou. Acquisition of data: Ohiorhenuan, Walker, Zhou, Godzik, Sagar, Farber. Analysis and interpretation of data: Ohiorhenuan, Zhou, Farber. Drafting the article: Ohiorhenuan, Walker, Zhou, Godzik, Sagar. Critically revising the article: Uribe, Walker, Farber. Reviewed submitted version of manuscript: Uribe. Statistical analysis: Ohiorhenuan, Zhou. Administrative/technical/material support: Sagar. Study supervision: Uribe.