Computer-aided designed 3D-printed polymeric scaffolds for personalized reconstruction of maxillary and mandibular defects: a proof-of-concept study

The primary objective of the study is to define the production process and verify its feasibility and reproducibility by simulating different scenarios of maxillofacial reconstructive surgery on cadaver heads.

Secondary objectives are: (a) verifying the accuracy of the reconstruction by comparing its morphological contour with the original one (bone morphological conformance); (b) evaluating the aesthetic outcome thanks to symmetry analysis with the contralateral side; (c) testing the efficiency and mechanical resistance of the reconstruction thanks to occlusal force tests; (d) conducting an exploratory test to include dental restoration into scaffold planning and printing.

The workflow was divided into 3 subtasks: (1) scaffold design and production; (2) simulation of the surgical procedure in a preclinical setting; (3) assessment of the reconstruction (Fig. 1).

We studied 6 types of bony defects (3 mandibular and 3 maxillary) with an increasing degree of complexity.

All maxillary defects had the medial bony cut on the midline of the hard palate, while they had a different vertical extension [11]:

For mandibular defects we referred to the classification proposed by Brown et al. [12]:

Computed tomography (CT) DICOM images were obtained for each cadaver head with the multidetector 128-slice CT scanner Somatom Definition Flash^® (Siemens AG, Munich, DE) with 0.7 mm axial slices and were subsequently uploaded to Materialise Mimics^®, version 21 (Materialise NV, Leuven, BE). The region of interest was segmented and the matching solid obtained was 'mirrored' on the contralateral side (Fig. 2). This step was meant to simulate the clinical condition where the ipsilateral bony anatomy is altered by the tumor, and the scaffold must be designed on the contralateral side. It was not performed for median defects (i.e., Brown type III). When present, the segmentation process also included teeth.

Next, cutting lines were created according to the abovementioned boundaries of the defects. The designed scaffold was exported in two.stl format files: one with cutting lines to be uploaded in the navigation system; the other without, for post-segmentation modifications (Meshmixer^®, version 3.5, Autodesk Inc, Mill Valley, US-CA and Solidworks^®, version 2019 Dessault Systèmes, Vélizy-Villacoublay, FR). Post-segmentation modifications depended on the type of the defect and included smoothening of the outer surfaces to ease printing and avoid injury to soft tissues, smooth cavitation of the maxillary scaffold (that was previously designed as a full solid), planning of holes for screws, and adding supports for scaffold fixation. The latter were tentatively applied only to mandibular scaffolds and were constituted by extensions of the inferior border of the scaffolds that were able to cover and adapt to the inferior border of the native mandible; they also had holes where the screws could be driven to fix the reconstruction (Supplementary Fig. 1). The designing process was finally checked through Gom Inspect^® (Zeiss, Jena, GE) to evaluate differences between the.stl file before and after the post-segmentation refinements; spatial discrepancies had to be < 1 mm.

The plug was printed in polylactic acid (PLA) (Pearl White PLA, Ø 2.85 mm, Ultimaker, Utrecht, NL) using Fused Filament Fabrication (FFF) technology with the 3D-printer Ultimaker 3 Extended^® (Ultimaker, Utrecht, NL). For each defect, 2 scaffolds were printed (one with the screw holes and one as a rescue option); an additional plug with supports for scaffold fixation was printed for mandibular defects (with supports for scaffold fixation).

Six fresh frozen cadaver heads (MedCure Inc, Portland, Oregon, USA) were dissected at the Dissection Laboratory of the University of Brescia, School of Medicine. The specimens (4 females, 2 males, mean age at death 71 years) had the arterial system injected with a bi-component red silicon (Down Corning, Midland, US-MI). In each cadaver head two defects were created on opposite sides (1 maxillary and 1 mandibular); therefore, each defect was tested twice. For occlusion analysis, the gingival-dental cast (Lascod S.P.A, Firenze, IT) of the upper and lower maxilla was taken before and after the procedure.

For maxillary resection, a Weber Ferguson lateral nasal skin incision with lip split was performed and a cheek flap was elevated to fully expose the bone. For mandibular resection, a cervical skin incision was performed and a visor flap was elevated without splitting the lip. Incisions and flap elevation were adapted according to the planned defect.

The extent of the resection was navigation-guided [13]. CT DICOM images with the designed scaffold and predicted cutting planes were uploaded to the Polaris Vicra^® optical navigation system (Northern Digital Incorporated, Waterloo, CA). Thanks to dedicated software (GTx-Eyes 3D navigation system^®, University Health Network, Toronto, CA), it was possible to navigate the CT on the surgical specimen and create the bony defect as planned and check the correct inclination of the saw, which is paramount to optimize the bone-scaffold interface (Fig. 3 and Supplementary Fig. 2). After this procedure, the scaffold was placed to fill the bony defect and fixed with 2.5 mm-diameter screws of different length (5–8–25 mm) and titanium plates (length × thickness: 31.5 mm × 1.5 mm; 26.2 mm × 1.5 mm; 31.2 mm × 1 mm). To test which scaffold worked best, the reconstruction was accomplished using first the scaffold with pre-designed holes for fixation with plates and screws. For the mandibular defects, the reconstruction was also repeated using the plug with supports for fixation (Fig. 4).

The quality and robustness of the reconstruction was tested through 3 different functional and quantitative tests.

The first parameter was the bone morphological conformance. The scaffold was segmented on postoperative CT DICOM images, and the surfaces of the native bone and the scaffold were superimposed and compared using Materialise 3-Matic^® software (Materialise NV, Leuven, BE). The measure of the mean morphological difference between the two surfaces was calculated through the root mean square deviation (RMS, Supplementary Fig. 3). This parameter has been used in different studies to compare the morphological conformance of maxillofacial reconstructions, and RMS less than 3 mm was considered good [14‐18]. We also performed symmetry analysis by comparing the defect side with the contralateral one. This was considered an aesthetic outcome.

The third test was a mechanical stress test to quantify the maximal bite force carried by the reconstruction until rupture of the plug or surrounding bones. For this purpose, we created a dedicated device that can measure the mechanical resistance of the scaffold with respect to a purely vertical force component generated by the operator closing the bite (Fig. 5). The tool was designed to test the vertical force generated on a single side of the oral cavity, so that the two reconstructions in the same cadaver head could be tested separately. For mandibular defects, the test was performed twice since two different types of scaffolds were used, and the fixation with screws and plates was always tested first. Subjective considerations of the operators were also annotated.

The production process required on average 40 h (range 36–44 h). In addition, the scaffold had to be soaked in water for 24 h to eliminate any traces of soluble components from the PLA printed device.

Overall, 30 scaffolds were printed: 12 for maxillary defects (6 with holes and 6 without), and 18 for mandibular defects (as above with the addition of 6 plugs with integrated supports for fixation). In all cases, dimensional gaps between the.stl file generated by the segmentation process and the.stl file obtained after post-segmentation modification were < 1 mm.

Twelve procedures (6 maxillectomies and 6 mandibulectomies) were performed on 6 cadaver heads (2 per head, 1 per side).

A navigation system was effectively set up in all cases. The navigation error was always < 1 mm, (mean: 0.65 mm; median: 0.62 mm). This accuracy was constant in both maxillectomies (mean: 0.69 mm; median: 0.65 mm) and mandibulectomies (mean: 0.60 mm; median: 0.54 mm). As per surgeon’s reports, positioning of the scaffold was always straightforward with an excellent bony-plug interface. In no case were adjustments of the bony edges deemed necessary after osteotomies were performed.

Scaffold fixation was efficient using screws and plates, and screws and PLA supports for mandibular plugs. In all cases fixation could be achieved with the scaffold with pre-designed holes for screws, and no drilling of new holes was required. In type III maxillary defects, the scaffold was further stabilized via a 25 mm-long screw driven in the pterygoid root under endoscopic guidance.

In relation to the mandibular scaffolds equipped with PLA supports, fixation was effective in all 6 cases and in no case did the support break during driving of the screws.

The procedure of scaffold placement and fixation was time saving. Overall, the mean time to fix the scaffold with plates and screws in all 12 defects was 50 min. The average time did not differ significantly between maxillary and mandibular plugs (51.6 and 48.3 min, respectively) with similar ranges of variations according to the complexity of the defect (40–60 and 35–55 min, respectively). Conversely, scaffolds with support for fixation (only for mandibular defects) were much quicker to fix and required only 14.5 min on average.

Maxillary scaffolds were tentatively resurfaced with vascularized tissue using temporo-parietal fascial flap or palatal mucosal island flap (oral side) and a naso-septal flap (nasal side). Coverage was overall satisfactory.

Data on bone morphological conformance, symmetry analysis, occlusal morphological conformance and strength tests detailed according to the defect and the single test are reported in Table 1.

In our algorithm, the role of the navigation system was paramount to obtain excellent correspondence between the predicted and real osteotomies, and optimize the matching between the surfaces of the scaffold and the bony edges, which is recognized as a key factor to allow new bone formation [24]. In our tests, the navigation system set-up was optimal in all cases (systematic error always < 1 mm); accordingly, the scaffold was always easily plugged into the defect and its surface perfectly faced the bony boundaries without any need for drilling to adjust the shape. This finding is in accordance with a preclinical study from our group on the use of 3D-printed scaffolds to reconstruct complex clival defects and highlight the necessity of a meticulous setting of the navigation system in a hypothetical clinical scenario [25].

Scaffold fixation was intuitive and timesaving. Interestingly, mandibular supports for fixation made the process much quicker and never broke either during driving of the screws or stress tests. Therefore, it is a promising technique to avoid the use of titanium plates, which are at risk of both extrusion and infection. In case of complex maxillary defects (type III), the positioning of a screw in the root of the pterygoid (if neither resected nor too pneumatized) can provide stability for the posterior part of the reconstruction. Even in this case, the use of the navigation system was essential to check for length and trajectory of the screw and avoid injury to the internal carotid artery.

Morphological conformances showed RMS values consistently below 3 mm (on average, in between 1.3 and 1.5 mm), which is considered optimal in the pertinent literature [14‐16, 26]. Reproducibility was apparently good, although it should be noticed that every procedure was repeated only twice.

Mechanical stress tests were successful, with maximal force values that were in line with the average human bite, ranging from 325 to 421 N [27‐29]. However, these results should be weighed cautiously, since the human bite requires the action of different powerful muscles that act along different vectors of force, which is impossible to recreate in a cadaver head. Our stress test was aimed at measuring the maximal vertical force carried by the reconstruction before breaking and could be only a faded surrogate of a human bite. It was mostly addressed to test the mechanical resistance of the fixation systems, which was overall satisfactory as the site of rupture never included this site.

With the aim of testing the efficacy of the algorithm in different scenarios, we analyzed three types of maxillary and mandibular defects with increasing degree of complexity from a reconstructive standpoint. Interestingly, our findings are apparently independent of the complexity of the defect (Table 1). This is relevant and supports the efficacy and applicability of this reconstructive protocol in different clinical scenarios.

Restoration of the dentition was an exploratory test to verify the accuracy of the production process in such a demanding setting. In an ideal clinical scenario, the production of a dentate scaffold would allow a definitive single-stage reconstruction. Overall, the task confirmed its complexity and made all the production process more cumbersome. In the future, more sophisticated protocols of scaffold planning and printing and different materials for teeth need to be investigated.

Our study has some limitations. First, the limited sample size prevented any statistical analysis, and the reconstruction algorithm could be tested only twice for each type of defect, which is the minimum to verify reproducibility. Then, we applied our protocol only in a preclinical setting on cadaver heads. This limits the validity of mechanical stress tests and precludes further analysis concerning the risk for infections, evaluation of new bone formation, and long-term integration of the scaffold.

As previously pointed out, the present study is intended as a preliminary feasibility analysis of a wider research line including bioengineering of the scaffold with hMSC. Therefore, the current algorithm is probably not applicable in clinics as it is.

Two relevant issues must be addressed: the protection from infections and the boost of scaffold integration.

The first one encompasses the resurfacing of the scaffold with vascularized tissue to avoid exposure to contaminated regions such as the oral, nasal, and pharyngeal cavities. In our study, this goal was attempted only for maxillary scaffolds using regional pedicled mucosal or fascial flaps. Coverage of mandibular plugs could be achieved by surrounding soft tissue in case of exclusive bony ablation, or by the fascio-cutaneous flap already used to restore larger soft tissue resections.

An alternative option could be the planning of a bi-phasic scaffold, with a rigid, resorbable core contoured by a cell-friendly bioactive shell to speed up mucosal coverage [30].

Scaffold integration has been investigated in animal models, such as rabbit [7, 20, 31] and dog [32] with uneventful follow-up and evidence of progressive mineralization of the scaffold. Han and coworkers described a small series of 3 patients with limited maxillary defects successfully repaired with patient-specific, CAD and 3D-printed polycaprolactone scaffolds that were inserted into the defect to recreate volume and achieve symmetry [33]. However, in case of extensive bony defects as the ones considered in the present study, scaffold integration through spontaneous bony repair is not conceivable. The bioengineering of the scaffold with hMSC is meant to speed up new bone formation, which ideally should replace in a timely manner the resorbable scaffold and recreate the ablated bone, providing a complete functional and aesthetic restoration of the defect. Preliminary results from our groups are promising [22, 23], although new tests in larger animal models are mandatory. In literature, so far several studies have investigated scaffold bioengineering with mesenchymal stem cells to propel new bone or cartilage formation [7, 34‐38]. Some attempts on large animals have yielded encouraging insights that bioengineered scaffolds also work in large size defects [39, 40].