Three-dimensional-printed custom-made hemipelvic endoprosthesis for primary malignancies involving acetabulum: the design solution and surgical techniques

Of all primary malignancies, 5–15% sarcomas occur in the pelvis, and chondrosarcoma, and Ewing's sarcoma along with osteosarcoma are the top three tumors [1, 2]. Besides, pelvis is reported as the third most common site of bone metastases following the spine and ribs [3]. Among pelvic tumors, tumors involving the acetabulum are closely related to poor hip joint function; therefore, how to treat them deserves special attention [4, 5].

Currently, en-bloc resection has been widely accepted, but correct and precise reconstruction of hemipelvic bone defect with low complication rate remains challenging. Previously, numerous reconstruction methods have been proposed, including iliofemoral arthrodesis or pseudarthrosis [6], massive allograft [7], and autoclaved autograft [8]; however, a large amount of studies revealed various limitations of these reconstruction options, such as highly occurred complications or poor functional results. In recent years, prosthetic reconstruction has predominated, on account of its easier approachability, better initial stability, more acceptable cosmesis, earlier weight bearing, and relatively rapid restoration of function [2].

In order to improve the implantation convenience and fixation stability, a lot of prostheses with different shapes and fixation methods has been reported, such as saddle endoprosthesis [1, 9], ice cream cone endoprosthesis [10, 11], modular hemipelvic endoprosthesis [2], and custom-made hemipelvic endoprosthesis [12‐14]. Nevertheless, bone implant integration is still unsatisfying in long-term follow-up due to outdated interfacial processing of the endoprosthesis and poor matching between endoprosthesis and residual host bone. To improve long-term survival of endoprosthesis, advanced technique for anatomy-imitating prostheses with a more efficiently osteoconductive structure is urgently demanded.

With understanding and progress on material science and three-dimensional (3D) printing technology (additive manufacturing), 3D-printed custom-made endoprosthesis with porous structure is considered to be a solution for these complicated and irregular bone defects, such as pelvic tumorous defect. To our best knowledge, 3D-printed custom-made hemipelvic endoprosthesis has already been reported with a promising result [15]. But previous reports mostly focused on clinical applications, the detailed design principles and methodology of this kind of endoprosthesis has not been described. Additionally, the worldwide criteria for 3D-printed endoprosthesis is still unestablished. Therefore, research and discussion on the design and methodology of 3D-printed endoprosthesis are inevitable.

Recently, we designed a kind of 3D-printed custom-made hemipelvic endoprosthesis and applied them in treating patients. Favorable clinical outcomes were observed. In this study, our aim is to describe the detailed design and surgical techniques of this endoprosthesis for hemipelvic bone defect.

Totally, 13 endoprostheses were designed on the basis of measured data. The mean tumor size was 7.7 (4 to 14) × 10.9 (6 to 18) × 8.5 (4 to 14) cm³ (Table 2).

As we have mentioned above, the endoprosthesis was composed of solid and porous structures. A 20-mm wide,and 5-mm thick solid titanium “arc-like” supporting structure was distributed from osteotomy plane after P1 resection, extended along arcuate line and ended at osteotomy plane after P2/P3 resection, connecting the other solid structures such as the screw holes, acetabulum, and pubis (Fig. 4). The porous structure was 600 μm pore size and 70% porosity. Averaged estimated resected volume endoprosthesis volume were 488.0 (186.7 to 732.8) and 155.9 ml (75.6 to 266.1). Averaged VPPS was 68.8% (53.0 to 86.0%), and mean endoprosthesis weight was 256.4 g (109.4 to 394.5). In cases whose iliac crest was reconstructed, small holes were distributed along the crest (Tables 2 and 3).

After P1 resection, fixation methods were determined by the volume of the residual ilium. Total resection of the ilium was performed in eight cases, endoprosthesis was directly fixed to the sacrum with a flat connecting surface. Orientation and number of ilio-sacral screws were designed according to the size of S₁ and S₂ vestibules. In three cases, most of the ilium was preserved, and endoprosthesis was fixed only to the ilium. Number and orientation of screws fixing in the iliac crest were modulated. Additionally, in two cases, most of the ilium was resected, the endoprosthesis was fixed to the sacrum by the same way as that after total resection of the ilium and to the residual ilium by screws (Table 3).

Therefore, in our series, prostheses were fixed to the sacrum in 10 cases. Mean area, length, and width of S₁ vestibule were 327.9 mm² (222.2 to 400.0), 23.1 mm (19.5 to 27), and 18.4 mm (15.8 to 21.0), respectively. The average area, length, and width of the S₂ vestibule were 131.7 mm² (102.6 to 147.6), 15.7 mm (13.9 to 17.3), and 10.6 mm (7.8 to 12.1), respectively. Median number of screws designed in the S₁ and S₂ vestibules was 5 (IQR, 4 to 6) and 1 (IQR, 1 to 2), respectively. In other three cases, endoprosthesis was fixed by radial screws inserting from the prosthetic acetabulum. Median number of screws designed in the ilium was 5 (IQR, 2 to 6; Tables 2 and 3).

The anteversion and inclination of the prosthetic acetabulum were reorientated: anteversion (anatomy vs. design, 20.2° ± 2.9° vs. 23.2° ± 1.7°, p = 0.013) and inclination (anatomy vs. design, 41.8° ± 2.3° vs. 42.4° ± 1.7°, p = 0.404) as seen in Tables 2 3, and 5.

After P2/P3 resection, fixation methods were determined by the preservation of the acetabulum and ischium. In three cases, the acetabulum was partially preserved, and solid acetabulum with porous layer was applied. In 10 cases, for the acetabulum that was totally resected even with part of the pubis, pubic endoprosthesis was created. For the residual superior pubic ramus or pubic symphysis with larger intermedullary width, the pubic stem was designed, otherwise, “cap-like” endoprosthesis with a tapered inner face wrapping pubic stump was created. Different screw fixations were applied according to preservation of the narrowest zone of pubis in residual hemipelvis to be fixed. If it was preserved, screws can be inserted from the endoprosthesis acetabulum to the residual pubis; otherwise, screws would enter from the pubic endoprosthesis (Table 3).

Ischial ramus reconstruction was determined by the preservation of the ischial tubercle. If so, the endoprosthesis imitating the contralateral ischial ramus was designed. In five cases whose ischial tubercle was resected, no reconstruction was performed (Table 3).

Averaged area, length, and width of NPSPMC were 79.8 mm² (40.4 to 126.2), 12.0 mm (8.6 to 16.3), and 8.3 mm (6.2 to 10.0), respectively. Median number of screws designed in the superior pubic ramus was 1 (IQR, 1 to 1). In 10 cases, whose endoprosthesis was fixed to the pubis, the average area, length, and width of RSSPMC were 200.5 mm² (103.8 to 333.2), 27.4 mm (14.2 to 41.1), and 8.8 mm (5.7 to 12.4), respectively. Fixation options were stem and “cap-like” structure in eight and two cases. The stems were 20-mm long, 12-mm wide, and 6.5-mm thick. The “cap-like” structures were 5-mm deep and owned the same size with the pubic resected surface at the rim of them. Ischial screw was designed in eight cases without total resection, and number was all two (Tables 2 and 3).

All of our osteotomy guides were fixed to the tumor side of the osteotomy plane. Ten millimeters was usually set as width of osteotomy guides, and holes for inserting Kirschner's wires were distributed along the middle line of osteotomy guides.

During surgery, median number of screws inserted in the S₁ vestibule was 4 (IQR, 3 to 4), in the S₂ vestibule was 1 (IQR, 1 to 1), in the ilium was 3 (IQR, 1 to 5), in the pubis was 1 (IQR, 0 to 1), and in the ischium was 1 (IQR, 1 to 1). Mean intraoperative time was 292.7 min (170 to 540), and blood loss was 3538.5 ml (900 to 8200). Precise implantation was observed: anteversion (design vs. post, 23.2° ± 1.7° vs. 24° ± 3.5°, p = 0.380) and inclination (design vs. post, 41.9° ± 1.7° vs. 42.9° ± 2.8°, p = 0.333). Postoperative pathological reports showed all tumors were removed as R0 resection (Tables 4 and 5).

Mean follow-up period of all the patients was 25.3 months (23 to 29). Averaged MSTS score was 22.1 (15 to 27) with a mean pain score of 3.7 (2.0 to 4.0), function score of 3.4 (2.0 to 4.0), emotional acceptance score of 4.2 (3.0 to 5.0), supports score of 3.6 (2.0 to 5.0), walking ability score of 3.9 (3.0 to 5.0), and gait score of 3.2 (2.0 to 4.0). Nine out of 13 require no supports, and four require a cane when walking a long distance. Mean ROM of hip joint was 119.2° (100° to 130°). No severe complications including infection, dislocation, aseptic loosening, and mechanical failure occurred (Fig. 9). T-SMART showed preliminary osseointegration three months postoperatively (Fig. 10, Table 4).

At present, prosthetic reconstruction is preferable for hemipelvic tumorous bone defect. A variety of prostheses have been developed since the 1970s [14], such as saddle endoprosthesis [1, 9], ice cream cone endoprosthesis [10, 11], modular hemipelvic endoprosthesis [2], and custom-made hemipelvic endoprosthesis [12, 14, 21, 22]. Despite providing acceptable clinical outcomes, endoprosthesis was associated with varies limitations. Saddle endoprosthesis and ice cream cone endoprosthesis require sufficient residual iliac bone. Besides, the saddle endoprosthesis and ice cream cone endoprosthesis cannot rebuild the intact pelvic ring, which results in the instability or looseness of the endoprosthesis. In recent years, modular hemipelvic endoprosthesis has been frequently used for different bone defects thanks to its high adaptability, nevertheless, intraoperative procedures are highly technically required to avoid erroneous implantation [23]. Custom-made hemipelvic endoprosthesis characterized by integrative design, has been reported with positive results [12, 14, 24]. No matter, manufactured by computerized numerical control technique, rapid casting technique, or 3D printing technology, these prostheses have a common shortcoming: absence of porous structure, resulting in poor osseointegration and inevitable mechanical failure [12, 14, 22]. Therefore, improving design and fabricating technique of custom-made endoprosthesis seems crucial.

With the development of 3D printing technology and profound understanding in osseointegration, novel 3D-printed custom-made endoprosthesis with porous structure has been reported with good early results [15, 23, 25‐27]. However, the details of endoprosthesis design was still unclear [27]. In our study, we firstly described the detailed design methodology step by step.

First of all, the feature of the preliminarily designed endoprosthesis needs modification. Simplification such as minimizing iliac wing and removing ischial spine and posterior iliac spines are good for wound healing, nerve protection and convenient implantation. Holes along endoprosthesis crest are for muscle reattaching, such as rectus femoris which is important for hip function. Again, the main body should be divided into solid and porous structures in order to reduce endoprosthesis weight, guarantee endoprosthesis strength, and enhance bone ingrowth [28]. Previous studies have reported that porous structure with pore size of 300 to 800 μm and porosity of 70% can enhance bone ingrowth [29‐33]. And post-processing procedure seems challenging in the porous structure under 600 μm-pore size [34]. Consequently, the porous structure with 600 μm pore size and 70% porosity was applied, and the follow-up outcomes showed well osseointegration in our patients. To achieve equal weight between resected specimen and endoprosthesis, the VPPS was modulated. Weight of resected specimen was calculated as product of its density and volume, and weight of endoprosthesis was calculated as total weight of porous and solid structures. The resected specimen is mixed by tumor, muscle, trabecular bone, and cortical bone so that its density was defined as the density of trabecular bone for tough estimation. Therefore, in our series, the VPPS was calculated as between 53 and 86%.

Integral configuration design was done, and then regional design was performed. Design after P1 resection plays the most important role [2, 16]. The major challenge is to fix the endoprosthesis rigidly. Long-term stability can be achieved by a well-designed porous structure, and initial stability is regarded as the fundamental factor. Firstly, flat connecting surface is required for tight connection and smooth intraoperative procedure. Secondly, all screws did not go through the sacroiliac joint, and the distribution of screws was determined by the residual volume of the ilium. Additionally, more and longer screws fixed in the S₁ vestibule were designed during preoperative simulation, and a back-up screw was designed in case of an unexpected situation [35]. Moreover, in the iliac screw design, radial distribution of screws was good for stress dispersion. As a result, in our follow-up, no mechanical failure was observed.

Design after P2 resection concentrates on the location and orientation of the prosthetic acetabulum which is crucial for hip joint function. The contralateral acetabulum was used for location imitation [15, 26]. Meanwhile, soft tissue removal and acetabular dysplasia correction were taken into consideration for acetabular orientation. Finally, thanks to our precise implantation, there were no significant differences between designed and postoperative acetabular location and orientation. And no dislocation was encountered.

In addition, reconstruction after P3 resection was performed to rebuild intact pelvic ring. Despite the extended plate being used assistantly for pubic endoprosthesis fixation in a previous study [21], it required more exposure as well as offered eccentric and poor force conduction. Therefore, central fixation of pubic endoprosthesis with specially designed connecting parts including stem or “cap-like” structure was designed basically according to the size of RSSPMC. Pubic stem is usually used in relatively larger pubis whose width of the resected surface is over 7.5 mm. But the “cap-like” structure is usually applied in thinner pubis whose width of resected surface is within 7.5 mm. In consideration of separating tendency after the destruction of pelvic ring integrity, assistant screws inserting to the residual pubic ramus are needed. The curvature of the superior pubic ramus reduces screw length in the bone; therefore, pubic screw is regulated by inserting from the acetabulum or pubic endoprosthesis for better fixation. With regard to ischial reconstruction, reports are quite limited. Our ischial design balances implantation difficulty and load transmission. Due to the deep location of the ischium, a back-up screw was applied to avoid fixation failure.

Intraoperatively precise osteotomy, proper endoprosthesis implantation, and correct screw insertion are critical in the operation of 3D-printed prosthetic reconstruction. Osteotomy guides were applied for providing desired osteotomy plane with less exposure and instrument requirement during operation [25, 26]. Additionally, due to the complex anatomy of the pelvis and severe displacement of the residual pelvic bone after resection, proper placement of endoprosthesis is technically demanding: (1) plastic endoprosthesis model can be used for the assessment of precise osteotomy; (2) conformation of well placement should be done before inserting screws by checking the acetabular orientation, pelvic continuity near osteotomy plane; (3) fixation between the sacrum or ilium and endoprosthesis is prior, then the reduction and fixation between superior pubic ramus and endoprosthesis; and (4) reduction using lag screw is required due to approachability of the ischium.

We also recognized limitations in this study. First of all, our short follow-up period might limit observation of unknown drawbacks. However, stable fixation, well osseointegration, and good functional outcome have been observed. Thus, long-term outcomes can be expected. Secondly, single uniform porous structure was applied. A porous structure with gradient porosity imitating the cortex bone and trabecular bone might be the best option for osseointegration; however, it is still challenging due to not necessarily uniform shrinkage. In addition, biomechanical analysis was not included in this study; therefore, finite element analysis and vitro tests focusing on (1) different porous structures, varying in pore size, porosity, and geometry configuration; (2) different VPPS; (3) different distributions of porous and solid structures; (4) different diameter, distribution and orientation of screws; (5) different fixation methods such as plate assisting fixation and lag screw fixation; and (6) different two pubic fixation methods in our design will be conducted to estimate the biomechanical performance and improve our endoprosthesis. Last but not least, although our series is one of the largest studies in 3D-printed custom-made hemipelvic endoprosthesis replacement, sample size with thirteen patients is still small which limited the power of the series. Therefore, a larger multi-institutional study is needed to ideally compare this approach with other types of reconstruction.

This study firstly presents detailed design methodology and related surgical techniques of 3D-printed custom-made hemipelvic endoprosthesis in treating malignancies involving acetabulum. Reconstruction with this kind of patient-specific implant is a multi-step process that involves measurement, design, manufacture, and surgery. Precise resection is performed with osteotomy guides and provided with foundation for complete reduction. Expectable bone ingrowth is conducted by highly matching surface, optimal porous structure, and well initial stability which is achieved by precise matching in 3D space including anatomical-conforming shape and proper fixation methods. Therefore, long-term outcome can be promising. Despite favorable outcomes, we observed some imperfections in the preoperative design and surgical application. More works are required in further study.