En bloc resection is a surgical technique that aims to achieve complete resection of vertebral tumors and provide adequate tumor margins.^1,2 En bloc resection can decrease the local recurrence rate and prolong survival in patients with aggressive benign or malignant spinal tumors.^1,3–5 However, restoring the structural stability of the spinal column after removal of the entire spinal segment is challenging.⁴ Given the extensive resection needed for aggressive benign or malignant spinal tumors involving the spinal segment, reliable anterior spinal reconstruction is of great importance for long-term stability.⁶ Although the clinical outcomes of traditional reconstruction techniques (including titanium mesh, expandable titanium, and stackable carbon cages, among others) after en bloc resection of thoracolumbar spine tumors are acceptable, instrumentation failure is not uncommon.⁴ Fusion rates following resection of spinal column tumors have been reported to vary from 36.0% to 100.0%, with rates of instrumentation failure after en bloc resection ranging from 0% to 40%.^7–11

A 3D-printed prosthesis, with its tailored shape, can be useful to reconstruct bone defects precisely and induce bone ingrowth through its innermost porous structure, which can help to reduce instrumentation-related complications. In recent years, this novel technology has gained popularity in spine surgery. Although 3D-printed prostheses are considered a promising option, most studies in this field are still limited to case reports or small sample studies with short-term follow-up. In our study, we investigated the outcomes of 23 patients who underwent en bloc resection for thoracolumbar tumors followed by spinal reconstruction using 3D-printed artificial vertebral bodies (AVBs) at > 2 years after surgery. The stability of this novel anterior reconstruction method was evaluated by CT, with Hounsfield unit (HU) values measured on CT used to assess fusion status.

Methods

Reconstruction

Two types of prostheses were used in our case series, a customized 3D-printed AVB and an off-the-shelf 3D-printed AVB (Fig. 1). For all patients in the series, no bone graft was packed within the 3D prostheses. The customized AVB was mostly used for reconstruction after multisegmental en bloc and/or sagittal resection. Prior to surgery in these cases, 1-mm thin-layer CT imaging of the target spine was performed. The DICOM data were imported into MIMICS software (version 15.0; Materialize) and Unigraphics NX (Siemens PLM Software) for prosthesis design. The anterior-posterior diameter and left-right diameter of both ends of the AVB were measured at a location about three-fourths of the adjacent endplates. The angle between the upper and lower ends of the prosthesis was adapted in situ to the adjacent vertebral bodies. Both ends of the prosthesis were slightly bulged to match the shape of the depression in the center of the contacting endplates. The porous surface and the innermost scaffold structure of the AVB, mimicking the native cancellous bone, were generated using a computer-aided design model for bony ingrowth and elastic modulus regulation. Each AVB prosthesis has a hole in the center that can be filled with bone graft; however, if not filled the hole in the center helps purge the residual titanium powder in the prosthesis, which is an important process in the fabrication of the prosthesis. The diameters of the pores and wires in the AVB were 600 ± 200 and 550 ± 200 μm, respectively, with an average porosity of 50%–80%. The nail buckle is installed into the ring hole of the 3D prosthesis, and the pedicle screw is installed in the ring nail buckle to complete the connection with the rod (Fig. 1E–I). The fixation structures were designed to improve primary stability and included screw holes or inner screw trajectories for connection to the adjacent vertebrae or a posterior fixation system per patient-specific needs. After careful consultation with the surgical team, the final design of the patient-specific AVB was sent to the AK Medical Company for prosthesis fabrication, which is performed using electron beam melting (Arcam EBM System) of Ti6Al4V powder. The above preparation process took approximately 7–14 days (Fig. 2).

Photographs of off-the-shelf and customized prostheses. A–D: Off-the-shelf prosthesis front (A), reverse (B), side (C), and transverse (D) sections. E–I: Customized prosthesis front section and front section of the nail buckle (E), side section of the nail buckle (F), reverse section (G), transverse section (H), and side section (I). Figure is available in color online only.

Manufacturing process of the prosthesis, including the close and frequent communication and cooperation between doctors and engineers required during the whole process. Figure is available in color online only.

Off-the-shelf AVBs, prefabricated using the electron beam melting system, are generally used for single-level reconstruction of the thoracic spine. These off-the-shelf AVBs are available in a variety of heights, ranging from 25 to 120 mm at a step difference of 2.5 mm. The endplates are arc shaped, with four different specifications (15 × 21, 18 × 24, 24 × 30, and 30 × 36 mm) and three angles (0°, 4°, and 8°). The off-the-shelf AVB has a porous surface and inner scaffold structure similar to that of the customized AVBs.

Posterior reconstruction was performed using pedicle screws, cobalt chromium rods (diameter 6.0 mm), and transverse connectors. Posterior instrumentation was adjusted to slightly compress the inserted AVB. The posterior elements were decorticated, and a morselized bone graft was placed around the intended fusion site. A protective brace was used in all cases for early mobilization, immediately after removal of the drainage tube.

Fusion

Fusion status was assessed using CT imaging at the final follow-up. A modified grading system was used to assess the titanium mesh cage fusion, per the methods described by Bridwell et al.¹³ A grade of I or II indicated achievement of anterior fusion, with grades III or IV indicating incomplete fusion. With regard to bone ingrowth, facilitated by the porous scaffold structure of the 3D-printed prosthetic implants, fusion was deemed successful when osseointegration of the bone-metal contact surfaces was achieved. However, it is difficult to grade bony ingrowth directly on imaging studies because of the dimensions of fenestrations.¹⁴ The value of the CT HU has been widely investigated as an alternative measure of the mineral density of spinal trabecular bone. In this study, we attempted to verify the density change inside the AVB by measuring the CT HU to provide in vivo evidence of bone ingrowth. The measurement method is as follows: the center images in the sagittal CT in PACS were obtained after the operation from each patient for measurement. The image was enlarged four times to facilitate measurement. The area was < 0.5 mm to the solid skeleton of the prosthesis near both ends of the endplate side and 0.5 mm away from the edge of the prosthesis side. A size range of 2 × 2 mm was taken from this area for CT HU measurement, which was performed by a surgeon with an independent radiologist. Four areas were measured on the upper and lower ends of the prosthesis and both sides of the prosthesis (Fig. 3). The average value was taken as the postoperative CT HU value of the patient.

Four areas were measured on the upper and lower ends and both sides of 3D prosthesis (green-outlined squares). Figure is available in color online only.

Results

Tumor-related death occurred in 2 patients with thoracolumbar spine tumors who underwent 3D prosthesis placement for anterior spinal reconstruction; 1 patient died 6 months after the operation and the other died 12 months after the operation. Four patients were alive at the time of this report but did not come to the hospital for follow-up imaging examination. Six patients were not enrolled because they did not have enough postoperative imaging data for this study. Therefore, a series of 23 consecutive patients (10 men and 13 women), with a mean age of 41 (range 17–71) years, who met the inclusion criteria, were enrolled. Of these patients, 18 had primary spinal tumors and 5 metastatic spinal tumors (Table 1). Moreover, 5 patients had a prior history of surgery at their local hospital. The median follow-up period after AVB implantation was 37 (range 24–58) months. Postsurgery, 14/23 patients were free of tumors, with local recurrence in 5, of whom 3 had systemic disease and 5 had systemic disease without local recurrence.

Treatment

En bloc resection was performed in 23 patients with thoracolumbar tumors, based on the WBB staging system. A type 2B resection was performed in 8 patients via a single posterior approach. A type 3B or 3C resection was performed in 7 patients and 3 patients, respectively, via an anterior-posterior approach; a type 7 resection was performed in 4 patients, via a posterior-anterior approach; and a type 5 resection was performed in 1 patient, via a posterior-anterior-posterior approach.

Single-level en bloc resection was performed in 12 patients, a 2-level resection in 3, a 3-level resection in 6, and a 4-level and 5-level resection in 1 patient each (Fig. 4). Total en bloc spondylectomy was performed in 20 patients and sagittal resection in 3 patients. The median operating time was 543 (range 241–987) minutes, with a median blood loss of 1200 (range 500–4300) ml.

Female patient aged 40 years with T8–10 chondrosarcoma. The patient was misdiagnosed with tuberculosis at her local hospital months prior and treated using posterior decompression surgery. Three months after this surgery, she was admitted to our hospital with the diagnosis of chondrosarcoma confirmed. The patient was treated with T8–10 en bloc spondylectomy using combined anterior-posterior approaches. A custom 3D vertebral prosthesis was implanted via an anterior approach and connected to the posterior screw rod system via a posterior approach. A and B: Preoperative anteroposterior lateral radiographs. C and D: Preoperative sagittal (C) and axial (D) CT images. E–G: Preoperative sagittal (E and F) and axial (G) MR images. H: Intraoperative view. I–L: Postoperative radiograph (I and J) and CT (K and L) images. Figure is available in color online only.

A total of 21 complications occurred in 18 patients, including pleural effusion in 6 patients, sensorimotor disorder in 5, CSF leakage in 4, wound infection in 3, venous thrombosis of the lower extremity in 1, heart failure in 1, and delirium in 1 patient. Postoperative radiotherapy was required in 12 patients and chemotherapy in 3 patients.

Reconstruction

A customized 3D-printed AVB was used in 10 patients, with an off-the-shelf 3D-printed AVB used in 13 patients. The AVB was implanted anteriorly in 5 patients and posteriorly in 18 patients. Definite fusion (grade I) was achieved in 10 patients (Fig. 5), probable fusion (grade II) in 10, probable nonunion (grade III) in 1, and definite nonunion (grade IV) in 2. The overall fusion rate (grades I and II) was 87%. Of note, fusion (grades I and II) was achieved in all 10 patients treated with the customized AVB (100%). Of the 13 patients in whom an off-the-shelf AVB was used, fusion was achieved in 10 patients (76.9%). The CT HU value for patients who achieved fusion (grades I and II) was significantly higher than the immediate postoperative value (1744 ± 321 HU) at 3 months (1930 ± 294 HU), 6 months (1997 ± 336 HU), and 1 year (1994 ± 257 HU) postoperatively (Table 2).

3D prosthesis 1 week after operation (A) and 2 years after operation (B). The visibility of the radiolucent line at the black arrows around the 3D prosthesis and the vertebral endplate decreases. It can be seen that new bone has grown in. The height of adjacent vertebral bodies in A and B is 68.98 mm, indicating that the prosthesis did not sink. Figure is available in color online only.

TABLE 2.

Postoperative AVB CT HU value differences

	Postop Period
	Immediate	3 mos	6 mos	1 yr
CT HU	1744 ± 321	1930 ± 294	1997 ± 336	1994 ± 257
p value	—	0.010	0.005	<0.001

Values are reported as mean ± SD.

There was no significant difference in prosthesis subsidence between the customized 3D-printed AVB and off-the-shelf 3D-printed AVB at 1 and 2 years postoperatively (Table 3). The average prosthesis subsidence at the final follow-up was 1.60 ± 1.79 mm, with a subsidence > 2 mm identified in 5 patients, as follows: 8.47 mm, 3.69 mm, 3.47 mm, 2.95 mm, and 2.42 mm. Of these 5 patients, 1 was treated using a customized 3D-printed AVB, with an off-the-shelf 3D-printed AVB used in the other 4. The patients with a subsidence of 8.47 mm and 3.69 mm experienced instrumentation failure. Chi-square test showed that there was no linear relationship between multilevel total spondylectomy (p = 0.69) or radiotherapy (p = 0.14) and prosthesis subsidence.

TABLE 3.

Differences in prosthesis subsidence for 3D-printed AVBs

	3D-Printed AVB		p Value
	Customized	Off the Shelf
Postop subsidence, mm
1 yr	0.93 ± 0.45	0.88 ± 0.84	0.869
2 yrs	1.08 ± 0.78	2.45 ± 2.69	0.315

Values are reported as mean ± SD.

Instrumentation-related complications, such as rod fracture, were observed in 2 patients at the final follow-up. Both of these patients were treated using an off-the-shelf 3D-printed AVB and fusion had not been attained. The rod fractures in these 2 patients were identified at the junction of rods and adjacent screws on the head side of the prosthesis. In 1 patient total vertebral resection in T11 was performed and osteotomy was performed in T10 and T12. Total vertebral resection of T12 was performed in another patient. Instrumentation failure occurred at 24 months postoperatively in one of these patients and at 36 months postoperatively in the other. Both patients underwent repeat revision surgeries because of repeated rod breakage.

Discussion

Our results show that short-term stability of anterior spine reconstruction after en bloc resection of thoracolumbar spine tumors is feasible using a 3D-printed AVB. The CT HU of bone-metal contact surfaces indicated bone ingrowth, with fusion achieved through the porous scaffold structure. To the best of our knowledge, this is one of the largest series to date on patients implanted with 3D-printed prostheses for anterior spinal reconstruction.

The 3D prosthesis shape can be customized, according to the individual needs of patients, with preliminary results indicating that it can be effectively used for spinal reconstruction.¹⁵ In 2016, the first clinical application of a 3D-printed vertebra was reported by our center in an adolescent boy who had a Ewing sarcoma of the second cervical vertebra.¹⁶ At the 1-year follow-up, the results showed no subsidence or displacement of the construct. Girolami et al.¹⁴ reported on the outcomes of 13 patients treated using a customized 3D-printed prosthesis for thoracolumbar spine reconstruction after en bloc resection for spinal tumors. At the average follow-up of 10 (range 2–16) months, one implant was removed following local recurrence of the disease and one revision surgery was performed for progressive distal junctional kyphosis. Wei et al.¹⁷ compared 3D-printed endoprostheses with two traditional reconstruction techniques after total en bloc sacrectomy; their findings suggested that the use of 3D-printed endoprostheses provided reliable spinopelvic stability. In agreement with these findings, our results showed that 3D-printed endoprostheses provided short-term thoracolumbar stability.

The attached fixation structure of the 3D-printed prosthesis can be connected to the posterior fixation system, which provides good immediate stability and a low rate of subsidence. The average prosthesis subsidence at the final follow-up in our case series was 1.90 ± 2.05 mm, with a subsidence > 2 mm identified in 5/23 patients. Of these 5 patients, a customized 3D-printed AVB was used in 1 patient and an off-the-shelf AVB in the other 4. The patients with substantial subsidence of 8.47 and 3.69 mm experienced instrumentation failure at 36 and 24 months after surgery, respectively. Both patients underwent revision surgeries for repeated rod breakage. In addition to internal fixation, prosthesis subsidence is related to many factors, such as multilevel total spondylectomy, postoperative radiation, and preoperative bone density. Further studies are needed to reveal the differences in the effects of 3D techniques on the mesh cage.

Various alternative devices for anterior column reconstruction have been described, such as bone grafts, mesh cages, and carbon fiber stackable cages, or expandable cages. The cost of bone graft is low and the potential for osteointegration is good; however, because of the long time required for the creeping substitution phase, an anterior plate is needed. Source material for large bone transplantation is limited, and allogeneic transplantation may lead to disease transmission. The use of mesh cages packed with cancellous bone graft is a common reconstruction technique that is a successful adjunct in restoring and maintaining sagittal plane alignment after thoracolumbar vertebrectomy.¹⁸ This combination has good potential for osteointegration; however, cage subsidence or instrumentation failure is a potential problem. The carbon fiber stackable cages are a better choice for a spinal tumor because they do not cause artifacts on imaging and thus enable early local recurrence detection. The main disadvantage of this method is the expense, which limits its use.¹⁹ The main advantage of the expandable cage is that these cages can be positioned from a minimally invasive approach and expanded in situ to the final size. However, an inadequacy of this method is the limited filling of bone grafts.²⁰

The use of 3D prostheses has had an increasing impact on precision medicine²¹ and provides an additional option for complex reconstructions.¹⁵ For complex spine oncology cases, the use of a 3D prosthesis allowed better preoperative planning, simplified the operative procedure, and enabled improved reconstructions, such as those involving more than three segments of total en bloc resection,¹⁴ which are difficult to accomplish with traditional titanium mesh.²² Anterior reconstruction using an AVB is indicated for patients in whom traditional spinal reconstruction techniques are not suitable, such as those requiring multilevel total en bloc spondylectomy.^23–26 In our case series, reconstruction after resection of 3 or more levels was performed in 11 cases (36.7%), including a 5-level reconstruction in a patient with a 26.8-cm bone defect. The slightly bulged contact surfaces of the AVBs matched the depression in the middle of the endplate of adjacent vertebrae to ensure uniform compression. Nail holes can be added laterally to the prosthesis. A good immediate stability of the prosthesis was achieved by connecting the nail hole screws with the posterior rods. Combined posterior reconstruction fixation to achieve reliable immediate stability allowed early postoperative mobility. In our case series, as soon as the drainage tube was removed patients were allowed to ambulate using a protection brace. In the series, an off-the-shelf 3D-printed AVB is used for single-segment vertebral reconstruction, and a customized 3D-printed AVB is generally used for multisegment vertebral reconstruction. Anterior and posterior insertion of the prostheses in the thoracic spine is easy because the bilateral nerve roots of the thoracic spine can be ligated. Moreover, a small, customized 3D prosthesis or an off-the-shelf prosthesis can be selected for placement from a posterior lumbar approach.

Achieving solid osseous fusion is crucial for maintaining long-term stability of spinal reconstruction. Currently, prefabricated prosthetic replacement devices, such as cages and AVBs, are commonly used for stabilization after en bloc resection of thoracolumbar spine tumors, with acceptable results for single-level vertebrectomy.²⁷ However, nonvascularized grafts may add to the rates of failure and nonunion, especially when adjuvant radiotherapy is required for patients with spinal tumors.^6,7,11,28 We found that long-term stability could be achieved with a 3D-printed prosthesis owing to bony ingrowth yielding an interlock of the bone-metal interface. In a previous study, we observed a preclinical correlation with histomorphometry and achieved osseointegration between the host bone and implants without bone grafting.²⁹ The open porous structure of the titanium alloy Ti6Al4V facilitated bone ingrowth, and the self-stabilizing AVB can maintain cervical spine stability in a sheep model.³⁰ Through superficial osseointegration of the adjacent vertebral bodies and establishment of the 3D prosthesis by ingrowth of bone through the porous structure on the surface of the prosthesis, in situ bone growth was finally achieved and the spine was stabilized anteriorly.³¹ Moreover, the modulus of activity of the prosthesis is enough to allow for strains that come back to the initial position after load release, while still providing a stable anterior mechanical support to the spine.¹⁴ Evidence of osseointegrative ability after upper cervical spine reconstruction has also been reported.^16,32 Wei et al.¹⁷ also found that osseointegration was facilitated at the bone-implant interface of 3D-printed endoprostheses. The radiolucent line around the 3D prosthesis and the vertebral endplate was fuzzy, which suggests osseointegration. To date, no accurate measure in determining bone ingrowth of the 3D prosthesis has been reported. This is the first observation of the CT-based measure of HU in determining bone ingrowth. Although bone ingrowth into the AVB is difficult to observe directly by using conventional imaging methods, owing to the dimensions of the fenestrations, the significant increase in CT HU values may indicate density changes within the trabecular pores. In our case series, the CT HU values within the AVB for patients in whom fusion was achieved (grades I and II) were higher than the immediate postoperative values and those at 3 months, 6 months, and 1 year postoperatively. The overall fusion rate in our case series was 85.7%. Probable nonunion was observed in 4 patients, with observable radiolucencies at the bone-metal surfaces. The CT HU increases were significantly lower in these 4 patients than in the patients who achieved fusion. These results indicate that CT HU measurement is a valuable method to detect osseointegration on the bone-metal interface of the 3D-printed prostheses. However, the standard deviation of the HU measured at the level of the implant was relatively high. The HU value in the 3D prosthesis will be affected by some factors, such as the influence of the solid skeleton of the prosthesis and the small measurement range. As more researchers participate in this research, the bias will be reduced and more valuable results may be produced.

Although our results are encouraging, this reconstruction method has some disadvantages which should be acknowledged. First, the reconstruction plan cannot be changed when a customized prosthesis is used. Second, the preoperative preparation of 3D prostheses is complicated and cumbersome because patient imaging information extraction, prosthesis design, and printing require a professional technical team and a certain amount of preparation time. An efficient and collaborative prosthetic design team is important, and precise communication with the surgeon is necessary. Third, in order to ensure successful placement of the 3D prosthesis during surgery, multiple prostheses with varying sizes need to be fabricated in advance, which increases healthcare economic costs. Finally, 3D prostheses cause radiographic artifacts like those observed with titanium cages. Therefore, CT and MRI metal artifact reduction techniques are needed to increase the visibility of periprosthetic anatomical structures to enable detection of tumor recurrence.

The limitations of our study must also be acknowledged. The group of patients is inhomogeneous in terms of expected ability to form new bone (due to age, nature of the lesion, previous radiation or chemotherapy, comorbidities, and bone quality). Although this is one of the largest oncological series of patients who underwent treatment with 3D-printed prostheses for anterior spinal reconstruction after en bloc resection of thoracolumbar spine tumors, larger samples, multicenter studies, and longer follow-up are still needed. In addition, in the current study there was no control group in the series, so the superiority of the technique we report over other reconstruction techniques could not be directly evaluated.

Conclusions

Anterior reconstruction using 3D-printed AVBs after en bloc resection of thoracolumbar spine tumors is feasible and reliable. The improved match of the prosthesis to the bone defect and attached fixation structures could provide good immediate stability. CT HU measurement can provide a valuable method to detect osseointegration on the bone-metal interface of the 3D-printed prostheses.

Acknowledgments

This work was supported by grants from Key Clinical Projects of Peking University Third Hospital (BYSY2017001).

We thank Editage for English-language editing.

Disclosures

The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper.

Author Contributions

Conception and design: Wei, Zhou, S Liu, X Liu, Z Liu. Acquisition of data: Wei, Zhou, Dang, Y Li, Zihe Li, Hu, Wang, Z Liu. Analysis and interpretation of data: Wei, Zhou, S Liu, Zhehuang Li, X Liu, Zihe Li, Hu, Z Liu. Drafting the article: Wei, Zhou, S Liu, X Liu, Y Li, Hu, Wang, Z Liu. Critically revising the article: Wei, Zhou, S Liu, Zhehuang Li, Dang, Zihe Li, Hu. Reviewed submitted version of manuscript: Wei, Zhou, Zhehuang Li, X Liu, Y Li, Hu, Wang, Z Liu. Approved the final version of the manuscript on behalf of all authors: Wei. Statistical analysis: Wei, Zhou, S Liu, Zhehuang Li, Zihe Li, Z Liu. Administrative/technical/material support: Wei, Zhou, X Liu, Dang, Hu, Wang. Study supervision: Wei, Zhou, S Liu, X Liu, Dang, Z Liu.