Revision surgical treatments and classification system for cage retropulsion in degenerative lumbar disease

Abstract

Background

Cage retropulsion is a severe complication of lumbar spinal fusion, causing neurological deficits and spinal instability. Current evidence on revision strategies is limited by small sample sizes, with no consensus on surgical planning.

Methods

This retrospective study analyzed 14 patients (11 males, 3 females; mean age 57.2 ± 11.2 years) who underwent revision surgery for cage retropulsion (> 3 mm posterior displacement) after lumbar fusion between 2013 and 2023. A novel 5-grade classification system (I–V) based on migration severity was proposed. Surgical approaches (anterior, n = 5; posterior, n = 9) were selected by revision timing, segment, and migration grade. Outcomes were assessed via VAS, ODI, and imaging; operative time, blood loss, and complications were recorded.

Results

85.7% (12/14) patients showed clinical improvement: VAS decreased from 7.3 ± 1.2 to 2.3 ± 1.1, ODI from 69.5 ± 12.4% to 15.2 ± 8.7%. The average operation time for anterior approach was 215.0 ± 44.5 min, and the average blood loss was 190.0 ± 156.2 ml. The average operation time for posterior approach was 281.9 ± 90.9 min, and the average blood loss was 387.5 ± 92.7 ml. 50% (7/14) had complications, mainly transient abdominal distension (n = 5). 85.7% (12/14) achieved fusion at 78.9-month follow-up.

Conclusion

This study proposes a practical classification system with preliminary feasibility in surgical approach selection for cage retropulsion. For symptomatic cases, early revision may yield better outcomes. Anterior surgery is recommended for L5-above segments with mild retropulsion, and posterior surgery is suggested for L5/S1 with severe retropulsion. In addition, anterior lumbar interbody fusion (ALIF) is a good option for L5/S1 with low-grade retropulsion when performed by experienced surgeon. Surgical success depends on tailored strategies considering revision timing, segments, and migration grade. But additional data are needed to confirm long-term safety and inform optimal management.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12891-026-09616-0.

Key points

1. We propose a five-level grading system for posterior cage migration, that shows preliminary potential for clinical application.

2. Owing to the anatomical characteristics of the lumbar vertebrae, significant displacement (Grades IV-V) is commonly seen at the L5/S1 segment, while it is rare in segments above L5.

3. The management of cage retropulsion should consider the time of displacement, involved segments, and degree of migration. Symptomatic cage retropulsion should be treated as soon as possible. Anterior surgery is suitable for segments above L5 with a lower degree of migration, whereas posterior surgery is appropriate for the L5/S1 segment with a more severe degree of migration. ALIF is also a reasonable alternative for carefully selected L5/S1 case with low-grade retropulsion when performed by experienced surgeons.

4. It is very necessary to completely remove the displaced cage.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12891-026-09616-0.

Introduction

Lumbar spinal fusion is widely recognized as an effective treatment for degenerative lumbar diseases, and lumbar spinal fusion using cages was first reported by Brantigan and Steffee in 1993s [1]. Compared with the allograft implant, the carbon fiber-reinforced polymer implant achieved a higher fusion rate. Many studies have proven the effectiveness of cages in promoting segmental stability, whereas some studies have reported the complications of using cages in lumbar spinal fusion [2–4]. Cage retropulsion, a rare but severe complication, is defined as the posterior translocation of the cage that exceeds 3 mm or extends beyond the wall of the vertebral body [5, 6]. Posteriorly migrated cage compresses the nerve roots or dural sac, which can cause severe back pain, radicular symptoms, or even cauda equina syndrome [7, 8]. Furthermore, cage retropulsion can lead to spinal deformity, narrowing of the disc space, and failure of intervertebral fusion [9]. High-risk factors for cage retropulsion, such as small cage size, screw loosening, and inadequate endplate preparation, have been widely reported in previous studies [5, 9–11]. These factors, which cause cage retropulsion, also lead to segmental instability and difficulties with vertebral fusion. Therefore, conservative treatment usually fails to improve the symptoms. Revision surgery for cage retropulsion is an effective but challenging procedure because of epidural adhesions and fibrous proliferation between the cage and vertebrae, as mentioned in almost all related studies [12–14]. The exposure of compressed nerve roots and posteriorly migrated cages is key point of surgery. During the first surgery, a part of the posterior bony structure was removed to achieve better decompression, and the gap left behind after decompression was filled with scar tissue, which further increased the difficulty of surgical exposure. There are currently some reports of successful cases of revision surgery for cage retropulsion, and a few studies have summarized the relevant surgical experience [15, 16]. However, the management of symptomatic cage retropulsion remains challenging, lacking a standardized decision-making framework for selecting the optimal surgical approach (anterior vs. posterior). Current evidence is limited to small case reports and surgeon preference, leading to inconsistent practices. Therefore, this study aimed to propose a novel, five-grade classification system for cage retropulsion, inspired by the Meyerding grading, with explicit incorporation of cage migration time, involved spinal segments, and migration severity as core classification dimensions. We hypothesized that by comprehensively integrating these key factors, the system could offer preliminary guidance for surgical approach selection in revision surgery. To explore the preliminary clinical feasibility and outcomes of this classification-based strategy, we analyzed what is to date the largest single-center cohort (n = 14) of patients undergoing revision for cage retropulsion, with the goal of providing initial insights for surgical planning in managing this complex complication.

Methods

Study design and patient selection

This was a retrospective, cohort study. A retrospective analysis was conducted on patients who underwent revision surgery for cage retropulsion after primary lumbar fusion between January 2013 and December 2023. During this period, 146 patients were admitted for lumbar revision surgery due to various causes, among whom 20 presented with cage retropulsion as the primary indication.

Per inclusion and exclusion criteria: 3 patients were excluded for incomplete follow-up/treatment at our institution; 1 for trauma-induced cage retropulsion; 1 for retropulsion secondary to progressive lumbar spondylolisthesis; 1 for asymptomatic Grade I retropulsion managed conservatively. Finally, 14 patients who met all inclusion criteria were included in the study cohort. The inclusion and exclusion criteria were operationalized and quantified to avoid subjective judgment bias. All patients underwent revision surgery via an anterior or posterior approach.

Inclusion criteria

1. Patients were initially diagnosed with lumbar intervertebral disc herniation, lumbar spinal stenosis, and other lumbar degenerative diseases before the first surgery, and treated with lumbar fusion surgery.

2. Radiographs obtained during follow-up confirmed cage retropulsion, and the patients had obvious clinical symptoms and underwent revision surgery with a follow-up of at least two years.

Exclusion criteria

1. Patients with congenital lumbar spine developmental abnormalities or deformities, or severe deformities of the surgical segment due to acquired conditions;

2. Primary surgery failure due to trauma, infection, tumor, or rheumatism and immune factors;

3. Inability to complete follow-up for various reasons.

Surgical decision-making protocol

To minimize indication bias and standardize treatment, a predefined surgical decision-making protocol was established based on the proposed classification system and the affected spinal level. Surgeries were performed by 3 spine surgeons (≥ 10 years of experience, trained in revision techniques):

1. For cases with cage migration was discovered within 3 months after the initial operation, an initial approach was preferred.

2. For Grade II-III retropulsion at levels above L5 (L2-L5), an anterior approach (Direct Lateral Interbody Fusion, DLIF) was preferred to avoid challenging posterior scar dissection and minimize nerve root manipulation.

3. For Grade IV-V retropulsion, or any grade at the L5-S1 level, a posterior approach (Transforaminal Lumbar Interbody Fusion, TLIF) was preferred, leveraging the relative spaciousness of the canal at L5-S1.

4. The final surgical approach was determined after comprehensive discussion of this protocol with the patient. This standardized process aimed to ensure consistent decision-making rather than reliance on individual surgeon preference.

Outcome measures

We evaluated patient characteristics such as age, body mass index, bone mineral density, diagnosis, and initial surgery details, including preoperative and postoperative symptoms, surgical segment, surgical method, cage type, and postoperative imaging manifestations. The details of revision surgery, including surgical approach, surgical segment, operation time, blood loss, preoperative and postoperative symptoms, and complications, were collected. Data (e.g., operative time, blood loss, complications and images) were sourced from anesthesia/surgical/nursing records and PACS. A third party verified 30% of records (error rate < 5%).

Clinical outcomes were evaluated using visual analog scores (VAS) and the Oswestry Disability Index (ODI). Blinded assessment and standardized measurement were implemented for outcome indicators. For clinical outcomes (VAS, ODI), two independent orthopedist (blinded to the surgical plan) collected scores before surgery, 3 months after surgery, and at the final follow-up (≥ 2 years) through telephone follow-up and outpatient re-examination. If the difference between two scores exceeded 1 point, the original follow-up records (e.g., patient diaries, electronic medical record score screenshots) were retrieved for review.

Imaging evaluation included radiograph, computed tomography (CT), and magnetic resonance imaging (MRI). Cage retropulsion is defined as a posterior translocation of the cage that exceeds 3 mm or extends beyond the wall of the vertebral body on the lateral X-ray image. For cage retropulsion (> 3 mm), two independent radiologists (blinded to the surgical plan) measured the distance between the posterior edge of the cage and the posterior wall of the vertebral body using lumbar lateral X-rays and CT sagittal reconstruction with ImageJ software. If the difference between two measurements exceeded 0.5 mm, a third radiologist conducted a review, with the goal of achieving an intraclass correlation coefficient (ICC) ≥ 0.85 for measurement consistency. The degree of cage retropulsion was defined as the ratio of the posterior displacement distance of the cage measured on the lateral radiography of the lumbar spine to the longitudinal diameter of the cage. The presence of a radiolucent area is thought to be an indicator of implant loosening on radiography, and screw loosening is defined using 1 mm as a width threshold [17]. Lumbar non-fusion or incomplete fusion includes the following three situations: the range of motion (ROM) of the lumbar segment is greater than 3°, there is a bright area in the fused intervertebral space on radiography, or there is no clear bony connection on radiography. Conversely, if there are no such manifestations, normal fusion of the lumbar vertebrae should be considered. A CT scan is used as a supplement to radiography for more precise measurements. Additionally, a CT scan was used to determine the direction of cage displacement in the horizontal plane and the degree of endplate damage. MRI was used to determine compression of the contents within the spinal canal. Imaging outcomes: fusion rate (modified Brantigan criteria) and screw loosening (lucent area > 1 mm, avoiding screw cap) were judged by 2 blinded spine surgeons (Kappa ≥ 0.8).

Statistical analysis

Descriptive statistics were used to summarize the data. Continuous variables ( e.g., VAS scores, ODI scores, operative time, and blood loss) were presented as mean ± standard deviation (SD). Categorical variables (e.g., fusion rates and, complication types) are reported as frequencies and percentages. Given the small sample size (n = 14) and descriptive nature of this study, no inferential statistical analyses (e.g., comparative tests) were performed. All the analyses were performed using Microsoft Excel (version 2019).

The grading of cage retropulsion

According to the posterior displacement distance of the cage, we propose the grading of cage retropulsion (Fig. 1): Grade I-posterior translocation of the cage exceeds 3 mm, but does not exceed the wall of the vertebral body; Grade II, the posteriorly migrated cage exceeds the wall of the vertebral body but < 25% of cage longitudinal length; Grade III, translocation distance > 25% of cage longitudinal length but < 50% of cage longitudinal length; Grade IV, translocation distance > 50% of cage longitudinal length, but < 75% of cage longitudinal length; Grade V, translocation distance > 75% of cage longitudinal length ( all the data measurements are based on the lumbar X-ray lateral views and horizontal CT scans).

Fig. 1.

The grading of cage retropulsion on the sagittal image (Grade I to V)

Results

The patients’ demographic and surgical results

A total of 14 patients with cage migration were included in this study (11 males and 3 females), with an average age of 57.2 ± 11.2 years. All 14 patients visited the hospital because of obvious low back or leg pain and were diagnosed with lumbar diseases (7 cases of lumbar disc herniation, 3 cases of lumbar spondylolisthesis, and 4 cases of lumbar spinal stenosis), all of whom underwent Transforaminal Lumbar Interbody Fusion (TLIF) surgery. Among them, six patients underwent multi-segment lumbar fusion, and eight patients underwent single-segment lumbar fusion. The fused segments included L2-S1. The average number of fused segments is 1.8. The cages used in the initial surgery included nano-hydroxyapatite (n-HA)/polyamide66(PA66), poly-ether-ether-ketone (PEEK), and titanium (Table 1). During the follow-up period, all the patients presented with low-back pain or radicular pain. After conservative treatment, the symptoms in these 14 patients did not improve significantly; therefore, revision surgery was performed.

Table 1.

Demographic and radiologic evaluation of the included patients

Fusion level
L2-L3	4
L3-L4	4
L4-L5	10
L5-S1	7
Retropulsion level
L2-L3	2
L3-L4	1
L4-L5	4
L5-S1	7
Fusion number
Single	8
Multiple(≥ 2)	6
Screw loosening(or fracture)	9
Cage type
Nano-hydroxyapatite(n-HA)/polyamide66(PA66)	2
Poly-ether-ether-ketone(PEEK)	7
Titanium	5

Imaging examinations were performed on all patients, and the imaging results showed that cage retropulsion occurred in 14 patients, including 2 cases in the L2-3 segment, 1 case in the L3-4 segment, 4 cases in the L4-5 segment, and 7 cases in the L5-S1 segment. CT tomography confirmed failed fusion at the posteriorly migrated segment in all patients, and 9 patients showed screw loosening or fracture at the failed fusion segment (Tables 1 and 2). Before the revision operation, except for 2 males with age younger than 40 years, all other patients underwent bone density tests by DXA or Quantitative Computed Tomography (QCT). Three patients were diagnosed with osteoporosis, with T values measured by Quantitative Computed Tomography (QCT) less than − 2.5 SD, and were treated with denosumab combined with oral calcium and vitamin D. Six patients were diagnosed with osteopenia, with T values ranging from − 2.5 to -1 SD. Oral calcium tablets and vitamin D were administered for treatment. Depending on the location of the failed segment and the degree of displacement, the anterior and posterior approaches were used to revise the cage retropulsion. All the patients underwent revision only at the failed segment (Table 2). The average operation time was 256.2 ± 83.1 min, and the average blood loss was 311.5 ± 154.6 ml. The average operation time for anterior approach was 215.0 ± 44.5 min, and the average blood loss was 190.0 ± 156.2 ml. The average operation time for posterior approach was 281.9 ± 90.9 min, and the average blood loss was 387.5 ± 92.7 ml (Table 3).

Table 2.

Characteristic of the patients who developed cage retropulsion (N = 14)

Case	Grade of CR	Age(yr)	Sex	Diagnosis	Fusion level	First fusion type	CR level	Time of CR (mon)	Cage materials
1	III	57	M	DDD	L2-L3	TLIF	L2/L3	48	NAK
2	II	57	M	DSL	L4-S1	TLIF	L5/S1	2	PEEK
3	IV	35	M	DSL	L5-S1	TLIF	L5/S1	3	PEEK
4	III	69	M	SS	L4-L5	TLIF	L4/L5	39	Titanium
5	IV	60	M	DDD	L4-S1	TLIF	L5/S1	3	NAK
6	II	27	M	DDD	L4-L5	TLIF	L4/L5	24	Titanium
7	II	63	F	SS	L2-L5	TLIF	L2/L3	4	Titanium
8	V	63	M	DDD	L5-S1	TLIF	L5/S1	12	PEEK
9	III	61	M	DDD	L4-L5	TLIF	L4/L5	4	PEEK
10	II	59	F	DDD	L5-S1	TLIF	L5-S1	12	Titanium
11	III	60	M	DSL	L3-L5	TLIF	L3-L4	3	PEEK
12	III	61	F	DDD	L4-L5	TLIF	L4-L5	48	PEEK
13	IV	64	M	SS	L2-S1	TLIF	L5-S1	16	Titanium
14	IV	65	M	SS	L2-S1	TLIF	L5-S1	48	PEEK

Case	Screw loosening	Revision surgery type	Revision surgery level	Symptom after CR	Postoperative symptom
1	Yes	TLIF	L2/L3	LBP+ radiculopathy	No
2	Yes	TLIF	L5/S1	LBP+ radiculopathy	LBP
3	No	TLIF	L5/S1	LBP+ radiculopathy	No
4	Yes	DLIF	L4/L5	Radiculopathy	No
5	Yes(fracture)	TLIF	L5/S1	Radiculopathy	No
6	Yes	DLIF	L4/L5	LBP	No
7	No	DLIF	L2/L3	LBP+ radiculopathy	No
8	No	TLIF	L5/S1	LBP+ radiculopathy + IC	No
9	Yes	PSR+DLIF	L4/L5	LBP+ radiculopathy	No
10	Yes	TLIF	L5-S1	LBP+ radiculopathy	No
11	Yes	TLIF	L3-L4	LBP+ radiculopathy	LBP
12	No	DLIF	L4-L5	LBP+ radiculopathy	No
13	No	TLIF	L5-S1	LBP+ radiculopathy + IC	LBP+ radiculopathy
14	Yes	TLIF	L5-S1	LBP + IC	No

CR cage retropulsion, M male, F female, DDD degenerative disc disease, DSL degenerative spondylolisthesis, SS spinal stenosis, TLIF transforaminal lumbar interbody fusion, PSR Posterior broken screw removal, DSC dural sac compression, PEEK poly-ether-ether-ketone, NAK nano-hydroxyapatite (n-HA)/polyamide66(PA66), DLIF Direct Lateral interbody Fusion, LBP low back pain, IC intermittent claudication, CES cauda equina syndrome

Table 3.

Characteristics of revision surgery

Surgery type
TLIF	9
DLIF	5
Surgery level
L2-L3	2
L3-L4	1
L4-L5	4
L5-S1	7
Duration of surgery (min)	256.2 ± 83.1
Intraoperative blood loss (ml)	311.5 ± 154.6
Dispose of migrated cage
remove	12
retention	2
Postoperative complication
unstable circulation and decreased oxygen saturation	1
alcohol withdrawal delirium	1
transient abdominal distension	5

TLIF transforaminal lumbar interbody fusion, DLIF Direct Lateral interbody Fusion

The surgical approach conformed to the predefined protocol in 14 cases. Among these 14 patients, 12 (85.7%) showed significant clinical improvement. The VAS scores decreased from an average of 7.3 points preoperatively to 2.3 points postoperatively, and the ODI scores decreased from an average of 69.5 points preoperatively to 15.2 points postoperatively. Notably, in one case (Case 9, L4/5 Grade III), initial posterior approach was attempted but abandoned due to excessive scar tissue and risk of nerve root injury; subsequent conversion to the protocol-recommended anterior approach (DLIF) led to successful cage removal and a good clinical outcome. The two patients whom the migrated cage could not be removed without clinical improvement (Case 2 and Case 13). The CT obtained during follow-up showed that the revision segments of 12 patients were normally fused, while 2 patients who underwent cage reinsertion because of difficulty in removing the cage showed poor fusion.

Complications were encountered in 7 patients (Table 3). A patient who underwent posterior approach surgery was admitted to the intensive care unit for 24 h and received ventilator support due to unstable circulation and decreased oxygen saturation. After symptomatic treatment, the patient’s condition returned to normal. One patient developed alcohol withdrawal delirium postoperatively due to long-term alcohol consumption. Five patients who underwent the anterior approach surgery experienced varying degrees of transient abdominal distension, and all symptoms were completely relieved before discharge. During the long-term follow-up of all the patients, no complications related to the wound or internal fixation devices occurred.

Case presentation

Case 1

A 63-year-old male patient with L5/S1 fusion cage retropulsion after L5-S1 TLIF surgery (Grade V).

Because of the L5/S1 intervertebral disc herniation, the patient underwent TLIF surgery at another hospital, and the symptoms were relieved after surgery. One year after the operation, the patient returned to the same hospital for removal of the lumbar implant. During the recovery period after implant removal surgery, the patient experienced lower back pain. One month later, the lower back pain worsened, accompanied by numbness in the right lower limb and intermittent claudication. The patient was then admitted to our hospital. Physical examination revealed tenderness at the L5-S1 segment, tense paravertebral muscles, and mild limitation of lumbar movement. The dorsiflexion muscle strength of the right foot was grade 4, and the dorsiflexion muscle strength of both big toes was grade 4. The right straight leg raising test result was positive, and the Babinski sign on both sides was negative. The imaging examination results indicated that the L5-S1 intervertebral fusion cage had significantly migrated posteriorly, with a posterior migration distance of over 75% of the cage longitudinal length. Therefore, we performed revision TLIF at L5-S1, removed the migrated cage, and re-implanted autologous bone and a new cage, followed by re-fusion and internal fixation at L5-S1. Postoperatively, the patient’s symptoms improved significantly. At the 2-year follow-up, imaging showed good fusion at the surgical segment (Fig. 2).

Fig. 2.

63 man, preoperative, postoperative and follow-up images. A-B The sagittal DR and sagittal CT of L5-S1 at 1-year follow-up, the retropulsion distance＞ 75% cage's longitudinal diameter (in sagittal image). D The axial CT of L5-S1 at 1 year follow-up. C and E The sagittal and axial MR imaging at 1-year follow-up, posteriorly migrated cage compressed the dural sac and right S1 nerve root, patient complained of low back pain and right lower extremity pain and intermittent claudication. Postoperative images of revision operation in follow-up (Posterior L5-S1 intervertebral cage removal, autologous bone and cage graft fusion and internal fixation surgery) (F-G) The sagittal DR and sagittal CT at 6 months after revision surgery, (H-I) The sagittal and sagittal CT at 48-months follow-up. The red line on the cage indicates the cage's longitudinal length and the distance it has migrated

Case 2

A 62-year-old female patient with L2/3 fusion cage retropulsion after L2-L5TLIF surgery (Grade II).

The patient was diagnosed with L2-5 spinal stenosis at another hospital because of pain and numbness in the left lower extremity and underwent L2-5 TLIF treatment. Three years after the surgery, the patient visited our outpatient department with, complaints of lower back pain for 9 months, which had worsened with left lower extremity pain and intermittent claudication for 1 month. Physical examination revealed tenderness in the lumbar region, accompanied by radiating pain in the left lower extremity, mainly in the left thigh and anterior knee. Skin sensation in the lateral aspect of both lower legs was slightly reduced. A partial limitation of lumbar movement was observed. The straight leg raising test was negative in both lower extremities. Reviewing the postoperative imaging data, lumbar DR and CT showed that the L2-3 intervertebral fusion cage was in the intervertebral space but slightly posterior. Lumbar CT obtained 3 months after the initial surgery indicated that the L2-3 intervertebral fusion cage had moved posteriorly, with a posterior displacement distance of less than 25% of the cage longitudinal length. After admission, DR, CT, and MRI examinations were completed, showing that the L2-3 fusion cage had shifted to the left posteriorly, with the posterior edge protruding into the spinal canal and compressing the left nerve root. Therefore, we decided to perform L2/3 DLIF, remove of the fusion cage, perform iliac bone graft fusion, and perform internal fixation. After the surgery, the patient’s symptoms completely improved. At the 72-months follow-up, imaging showed good fusion at the surgical segment (Fig. 3).

Fig. 3.

62 woman, preoperative, postoperative and follow-up images. A The sagittal DR at 3 months follow-up showed L2-3 cage had translocated posteriorly, and the retropulsion distance ＜25% cage's longitudinal diameter. B and D Sagittal and axial CT of L2-3 at 27 months follow-up, (C and E) The sagittal and axial MR images at 27 months follow-up showed posteriorly migrated cage compressed the dural sac and left L3 nerve root, and the patient complained of low back pain and left lower extremity pain and intermittent claudication. F-G Sagittal DR and sagittal CT at 3 days after revision surgery, (H-J) Anteroposterior, sagittal DR and sagittal CT at 3 years follow-up. The red line on the cage indicates the cage's long diameter and the distance it has migrated

Discussion

This study presents the largest known cohort of patients undergoing revision surgery for cage retropulsion. Its primary contribution is the introduction and preliminary clinical feasibility assessment of a novel classification system, which aims to provide a reference for standardizing surgical approaches selection in an area previously guided largely by individual experience. After reviewing previous case reports and our own cases, we found that when patients develop symptoms due to cage retropulsion, the symptoms are difficult to improve through conservative treatment, and may even gradually worsen. Therefore, once cage retropulsion is confirmed on radiography, and corresponding symptoms are present, the early surgical treatment should be recommended. Conversely, conservative treatment can be chosen for asymptomatic patients. In addition to the above cases, one patient was found to have cage retropulsion (grade I). The patient was asymptomatic; therefore, conservative treatment and regular follow-up were advised.

The approach to revision surgery has been controversial in the previous studies. There was a report of successful revision surgery 20 years ago by McAfee et al. [18]. They successfully extracted 4 cages that had migrated into the spinal canal and caused iatrogenic spinal stenosis through the posterior approach. Successful cases using the posterior approach have also been recently reported [7, 19]. Anterior techniques have also been used to treat cage retropulsion, with satisfactory results reported in some studies [12, 20, 21]. Despite these successful cases, there is no clear standard for choosing the surgical method. To the best of our knowledge, only one study has reported on the design of a surgical plan. Tanaka et al. [14] documented three patient types affected by cage retropulsion. Type one patient experienced only low back pain without radicular symptoms and were treated with minimally invasive TLIF from the contralateral side without cage removal. Type two patient, who experienced both low back pain and radicular symptoms, were successfully treated with PLIF to remove the migrated cages. Type three patient had cauda equina syndrome or severe radicular symptoms, and anterior cage removal was recommended. This study comprehensively summarizes the treatment plans for cage retropulsion and, provides a perspective for the surgical treatment of cage retropulsion. This study prompted us to consider the design of appropriate surgical plans. Based on the study of our cases and previous research, we found three key factors, including revision surgery timing, cage retropulsion level, and the extent of retropulsion of the cage, have a significant impact on the selection of surgical plans.

According to previous studies, posterior displacement of the cage usually occurs within three months after surgery [5, 14]. Scar hyperplasia of the dura mater usually occurs three months after the first surgery. Cage retropulsion may lead to the aggravation of epidural adhesions and fibrous proliferation in the subsequent period, increasing the difficulty of revision surgery. Based on previous studies and our experience, cage retropulsion should be diagnosed and treated immediately. If revision surgery is performed using the original surgical approach within 1 month after the initial surgery, before the formation of dural adhesions and scar hyperplasia, a high probability of success can be expected without serious difficulty or risk. If revision surgery is performed through the original surgical approach within 1 to 3 months after the initial surgery, successful removal of the cage can be achieved in most cases; however, in a very small number of cases, it may be difficult to expose the cages because of adhesion between the scar and neural structures, which may induce cage removal failure. Among our surgical cases, there were 4 cases with cage migration was discovered within 3 months after the initial operation and revision surgery was performed. All 4 patients underwent revision surgery using the original approach. In three cases, the cages were successfully removed and replaced, but in one case (L5/S1), cage removal was difficult. Therefore, the cage was struck forward to avoid compression of the dura mater and the nerve roots. If revision surgery is attempted through the original approach 3 months after the initial surgery, the surgical difficulty will increase significantly due to the formation of fibrous scars [22–24], and more patients may encounter failure of cage removal. For such patients, revision surgery through the original approach may not be the best choice, and the surgical approach should be selected based on the affected segment and degree of cage retropulsion.

According to the grading of cage retropulsion, there were four patients with degree II, five patients with degree III, four patients with degree IV, and one patient with degree V. In our cases, for the cage retropulsion segments at L2-L5, the degree of displacement did not exceed grade III. However, for the L5/S1 segment displacement cases, the degree of displacement was usually more severe (grade IV or V). We speculate that this is mainly because the L5/S1 segment is particularly unique, as a significant portion of cage retropulsion occurs in this segment, which may be related to local anatomical structures or local biomechanical properties. Compared with the other nerve roots, the S1 nerve root emerges from the dural sac earlier and has a longer course; therefore, it has greater freedom of movement within the spinal canal [25, 26]. Therefore, if cage retropulsion occur at L5/S1, the blocking effect of the dural sac and nerve root on the cage is smaller than that of the other lumbar segments, and patients usually do not experience significant symptoms until the cage has significantly shifted posteriorly. Conversely, high-grade cage retropulsion is less likely to occur above the L5 segment because of the blocking effect of the dural sac and nerve root.

For cage retropulsion above the L5 segment, the posterior approach would face difficulty exposing the cage. Meanwhile, owing to the high tension of nerve roots in these segments, it is also challenging to remove the cage without injuring the nerve roots through the posterior approach. Compared to the posterior approach, it is much easier to expose the posteriorly migrated cage through the anterior approach. Although there are some challenges in catching and removing the cage through the anterior approach, compared to the challenge of the posterior approach, it is acceptable. Therefore, we believe that the anterior approach surgery may be more suitable for these patients. In our cases, five patients with cage retropulsion above the L5 segment ( two grade II, three grade III) underwent revision surgery through the anterior approach. All cages were successfully removed, and new larger cages or iliac bone blocks were implanted. All 5 patients recovered well after revision surgery, and the symptoms improved significantly. In our practice, we also attempted to perform revision surgery through the posterior approach for patients with cage retropulsion above the L5 segment (2 cases of grade III and 1 case of grade II). However, the revision surgeries were quite challenging, especially during the process of exposing the cage, because the nerve roots are stuck by scars, are not easy to recognize and separate, and must be pulled hard to expose the cage, which may cause injury to the nerve roots. For one patient (case 9, L4/5, grade III) with cage retropulsion and screw fracture, we initially attempted a posterior approach for revision in order to replace the screw and cage simultaneously; however, during surgery, we found it was too hard to expose the cage, although the L5 nerve root had been pulled very hard, and concerns about nerve root injury forced us to switch to the DLIF approach. Eventually, we successfully removed the cage using the anterior approach and achieved good surgical outcomes. Some inherent advantages of the anterior approach have also been reported in many studies. Shorter operative time, reduced blood loss, smaller wounds, prevention of adjacent segment degeneration, disc height restoration, and high fusion rates have been mentioned as advantages of anterior approach operation [27–30]. Therefore, for cases with cage retropulsion of a low-grade (grade II or III) above the L5 segment, anterior surgery should be prioritized. For patients with high-grade (grade IV or V) cage retropulsion above the L5 segment, it may not be easy to catch and remove the cage through the anterior approach. However, as mentioned above, high-grade cage retropulsion is less likely to occur above the L5 segment, and in our cases, there was no patient with high-grade cage retropulsion above the L5 segment. Therefore, this concern may have been unnecessary.

As mentioned earlier, the L5/S1 segment is rather special, and the degree of cage retropulsion in this segment is generally more severe. However, simultaneously, it is easier to expose and remove the cage during a posterior approach. Additionally, the risk of nerve root injury was lower in this segment than in other segments. Patients with a high iliac crest and L5/S1 level surgery should also be considered contraindications for anterior approach surgery [12]. Moreover, screw loosening is common in patients with cage retropulsion, and screw fractures are also common in some patients. Some studies have found that screw loosening is a risk factor for cage retropulsion [8, 9, 31]. Therefore, we believe that renovating the posterior fixation system or adding an additional fixation system is important for achieving long-term stability. In our study, nine patients (two with grade II, two with grade III, four with grade IV, and one with grade V) eventually underwent revision through the posterior approach. Among them, there were seven cases at the L5/S1 segment (two with grade II, one with grade III, four with grade IV, and one with grade V), one case at L2/3 (grade III), and one case at L3/4 (grade III). For patients with cage retropulsion at the L5/S1 segment and with a high grade of cage retropulsion, although posterior surgery also has difficulty separating and exposing the cage, the risk of nerve injury is much lower than that of the segments above L5, and the success rate of cage removal is quite high. Although the posterior approach (TLIF) remains the preferred strategy for most high-grade (IV-V) retropulsion at the L5/S1 segment, anterior approaches such as ALIF may be a reasonable alternative for carefully selected low-grade (II-III) cases when performed by surgeons with extensive experience. Previous studies have reported successful revision outcomes using ALIF for low-grade cage retropulsion at L5/S1, which supports its role as a complementary option in clinical practice [18, 21].

Whether the anterior or posterior approach surgery is adopted, can the posteriorly migrated cage be left in place without removal? As mentioned in the research by Tanaka et al. [14], type one patient experienced only low back pain without radicular symptoms and were treated with minimally invasive TLIF from the contralateral side without cage removal. The patient achieved good therapeutic outcomes after revision surgery. However, among our patients, 2 adopted the strategy of not removing the cage but reinserting it. Nevertheless, in the postoperative follow-up, the improvement in the patients’ symptoms was not satisfactory, and the images also showed that the segment was not fully fused. This might be because reinserting the cage cannot provide sufficient segmental stability, and there is a fibrous tissue barrier between the cage surface and the endplate surface, preventing the cage from achieving segmental fusion. Therefore, simply striking the posteriorly migrated cage back into the intervertebral space may not completely solve the cage retropulsion problem. It is best to remove the cage and replace it with a larger cage or an iliac bone block. In addition, fetching the cages presents a challenge. Under favorable exposure conditions, most of the cages were removed. For fragile cages that are difficult to remove, using a high-speed burr or osteotomy to break them apart can reduce the risk of dural sac injury. Additionally, some tools have been reported to effectively remove the cages [18, 27]; However, after partial removal of the cage, the remaining part may cause residual symptoms. Long-term research and practice are needed to explore the complete removal of the cages.

Literature has extensively documented the risk factors associated with cage retropulsion. These risk factors can be broadly categorized into demographic-, cage-related-, and surgery-related factors [5, 6, 8, 9, 11, 14, 32]. Among the numerous risk factors, some may not only be related to the failure of the initial surgery, but could also affect the success rate of revision surgery. In our cases, nine patients had varying degrees of low Bone Mineral Density (BMD) and, even osteoporosis. The fact that low BMD can affect the success rate of surgery has been reported in many clinical and biomechanical studies [8, 9, 33, 34]. In addition, the size and placement of the cages were important factors. Small-sized cages and posterior cage placement have been recognized by most studies as high-risk factors for cage retropulsion [9, 11, 35–37]. Insufficient posterior fixation instruments and endplate preparation for cage retropulsion have also been documented [5, 31, 38]. Based on the above high-risk factors, we suggest that the following points should be noted before and during revision surgery: (i) For patients with low BMD or osteoporosis, routine anti-osteoporosis treatment is required before and after surgery. According to the patients’ clinical and economic situation, a combination or sequential therapy of calcium supplements and denosumab, bisphosphonates, teriparatide, etc., can be selected [39]; (ii) in revision surgery, it is necessary to remove and replace the failed cages. The new cages should be placed slightly anterior to the center of the intervertebral space. (iii) Before placing the cage, the endplates need to be adequately prepared, but the bony endplates must not be damaged to prevent cage subsidence. Adequate bone grafting is necessary if the endplate is damaged. (iv) The compression of posterior fixation instruments should be adequate. If the screw fixation strength is insufficient, autogenous/artificial bone may be filled into the screw trajectory, and pedicle screws may be augmented with bone cement.

Limitation

Although this retrospective study provided important data, it had some limitations. First, although this study included 14 cases, because of the low incidence, the total number of migration cases was insufficient to draw any conclusions. Despite being the largest cohort to date, our sample size limited multivariate risk factor analysis. As some patients underwent their initial surgeries in other hospitals, it was impossible to obtain complete medical records and imaging data. Therefore, there might be deviations in factors such as the information of the initial surgery, time of cage retropulsion, and degree of retropulsion. More cases and studies are needed to clarify the formulation of the surgical strategy and optimization of the surgical procedure for cage retropulsion.

Although the interobserver reliability for cage displacement measurement (the key basis for grading) was satisfactory, the reproducibility of the 5-grade classification system itself has not been formally assessed. Specifically, we did not evaluate whether different observers would assign the same grade (I–V) based on the same displacement data, nor did we verify the consistency of a single observer’s grading results at different times. Future multi-center studies with more observers and a larger patient cohort are needed to confirm this consistency before the system is widely used clinically.

We acknowledge the limitations inherent in our study’s retrospective design and small sample size (n = 14). These factors limit the statistical power for comparative analysis and increase susceptibility to biases. However, the rarity of symptomatic cage retropulsion makes large, prospective cohorts impractical. In this context, the value of our study shifts from statistical generalization to proof-of-concept. The high rate of conformance to the protocol and the corresponding excellent clinical outcomes (85.7% improvement in conformant cases) provide compelling preliminary evidence for the clinical utility and feasibility of the proposed classification system. The single case where intraoperative conversion to the protocol-recommended approach was necessary (Case 9) ironically strengthens our thesis by demonstrating the challenges of deviating from the framework.

Conclusion

In summary, revision surgery for cage retropulsion is challenging in clinical practice, and careful selection and implementation of surgical plans are vital. For symptomatic patients, early revision surgery may yield better clinical outcomes. The timing of cage retropulsion, involved segments, and degree of retropulsion should be comprehensively considered to determine the choice of anterior or posterior approach. Anterior surgery is suitable for segments above L5 with a lower degree of migration, whereas posterior surgery is appropriate for the L5/S1 segment with a more severe degree of migration. ALIF is also a reasonable alternative for carefully selected L5/S1 case with low-grade retropulsion when performed by experienced surgeons. Notably, 50% (7/14) of patients had transient, non-life-threatening postoperative complications (e.g., abdominal distension, n = 5). Despite this, 85.7% (12/14) showed marked improvement (VAS: 7.3 ± 1.2 to 2.3 ± 1.1; ODI: 69.5 ± 12.4% to 15.2 ± 8.7%) and successful fusion (mean follow-up 78.9 months), supporting the preliminary safety and efficacy of revision surgery. No severe complications such as nerve injury occurred in this study, but such complications can significantly impair prognosis. Thus, additional clinical practice is needed to confirm long-term safety and effectiveness and provide further guidance for managing cage retropulsion.

Abbreviations

VAS

Visual analog scale

ODI

Oswestry Disability Index

ALIF

Anterior Lumbar Interbody Fusion

CT

Computed Tomography

MRI

Magnetic resonance imaging

ICC

Intraclass correlation coefficient

ROM

Range of motion

PACS

Picture Archiving and Communication System

SD

Standard deviation

TLIF

Transforaminal lumbar interbody fusion

n-HA

Nano-hydroxyapatite

PA66

Polyamide 66

PEEK

Poly-ether-ether-ketone

DXA

Dual-energy X-ray Absorptiometry

QCT

Quantitative Computed Tomography

DLIF

Direct Lateral interbody Fusion

CR

Cage retropulsion

M

Male

F

Female

DDD

Degenerative disc disease

DSL

Degenerative spondylolisthesis

SS

Spinal stenosis

PSR

Posterior broken screw removal

NAK

Nano-hydroxyapatite (n-HA)/polyamide66(PA66)

LBP

Low back pain

IC

Iintermittent claudication

CES

Cauda equina syndrome

DR

Digital Radiography

BMD

Bone Mineral Density

Authors’ contributions

All the authors contributed to the study conception and design of the study. The material preparation and analysis were performed by Zhao, Zheng, and Zhou. The first draft of the manuscript was written by Weiming Zhao, and all the authors commented on the previous versions of the manuscript. All authors have read and approved the final manuscript.

Funding

No funding was obtained for the conduct of this study and the preparation of the manuscript. All authors declare no competing interests, financial or non-financial, and have no relevant employment, financial interests or other relationships that could inappropriately influence this work. All authors confirm that there are no undisclosed conflicts of interest, and the study has been conducted in accordance with the principles of good clinical practice.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

As this was a retrospective cohort study, it was approved by the Ethics Committee of West China Hospital, Sichuan University, and informed consent was waived. Written informed consent was obtained from all participants for the use of their clinical data, intraoperative images, and radiological examinations for research and publication.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.