This study was reviewed and approved by the Research Ethics Committee of Chongqing Medical University, reference 2021-048. Informed consent was obtained from each participant. The validation study was conducted at a single academic training center.

As computed tomography (CT) scans are able to show bony structures precisely [28], this modality was selected as the imaging resource for reconstructing the 3D structure of the hip joint. The deidentified CT data of the right hip joint from a 31-year-old female volunteer were used. The alpha angle and lateral center edge angle for the volunteer’s hip joint were 60.9° and 35.4°, respectively. The CT scanning (BrightSpeed, 16-slice CT scanner; General Electric, Boston, MA) was conducted in the horizontal plane with a slice width of 0.35 mm. The CT scans were saved as DICOM format data using InVesalius software, version 3.1.1 (Renato Archer Information Technology Center, Campinas, Brazil), to reconstruct 3D volumetric data, which were imported into a CAD file with the STereoLithography format. Next, the CAD file was imported into Meshmixer software, version 3.5 (Autodesk, San Rafael, CA) for 3D segmentation, which removes unwanted structures in the image. In the process of segmentation, only a minimum smoothing algorithm was applied to retain the anatomical features of the scanned region.

For the design process, the CAD file was imported into Rhino software, version 6 (Rhinoceros, Seattle, WA). During this stage, unnecessary bony structures were removed, retaining only the anterior superior iliac spine, the acetabulum, and the proximal end of the femur. Since the volunteer was CT scanned in a normal lying position, the femoral structure was repositioned to simulate the hip in the tractional condition to enable access to the central compartment for the simulated operation [23, 28]. The box-shaped simulator was designed on the basis of a modular concept and comprised two main parts, one part as the soft component to simulate soft tissues and the other part as the hard component to simulate bony structures (Fig. 1A).

Design of the simulator. A Main components of the simulator: (a). clip frame to fix the soft component, (b). soft component to simulate the soft tissue, (c). container for holding all the bony and supporting structures, (d). bench fixer used in tandem with a bench vise to secure the simulator onto the table (foldable); B The 3D printed component of the acetabulum; C The 3D printed bony structures of the simulator; D Assembled simulator with the silicon component

The structural framework of the anterior superior iliac spine, the acetabulum, and the proximal end of the femur were designed as independent modules, which allowed them to be replaced when simulating various hip conditions. The hip joint capsule was designed with a water drop shape, gradually enlarging in form from the femoral head to the acetabulum. This would coordinate with the traction force applied to the hip joint. As the acetabular labrum was not clearly defined in the CT scan, it was designed manually to have a 3–4 mm thickness with a width of approximately 8 mm. Nine fixed markings were incorporated on the surface of the acetabulum from the 8 o’clock position to the 4 o’clock position to aid the intra-articular identification of anatomical structures (Fig. 1B).

The main body of the simulator, including the anterior superior iliac spine, the proximal end of the femur, and the container with all the structural components, was 3D printed using an EOS Formiga P100 3D printer (EOS GmbH Electro Optical Systems, Krailling, Germany) and fabricated with white polyamide EOS PA2200 (Fig. 1C). The components of the acetabulum and the acetabular labrum were 3D printed as a single piece using a Stratasys J750 3D printer (Stratasys, Minneapolis, MN) and fabricated using VeroPureWhite, VeroCyanV, VeroMagentaV, VeroYell and Agilus materials. The soft component was fabricated in a silicon material, Ecoflex 00-30 (Smooth-on, Macungie, PA), using a mold that was designed and 3D printed (Fig. 1D).

The simulator that was developed could be secured onto the table by using two bench vises. The materials of the simulator were photosensitive to radiography, and all the internal structures were visible by fluoroscopy (Fig. 2A). The anatomical landmarks of the anterior superior iliac spine and the greater trochanter could be identified by feeling the simulated soft tissue. When viewed using the arthroscopic camera, the intra-articular anatomical structures of the simulator would appear similar to those observed in the hip joint of the human body (Fig. 2B).

Usage of the 3D printed simulator. A Fluoroscopic images for the portal placement process; B Arthroscopic view of the process for establishing the distal anterolateral (DALA) portal

To evaluate the validity of the simulator as an assessment tool, a cross-sectional study was conducted. Twenty-nine orthopedic surgeons from 11 hospitals were invited to participate in the validation study. All the participants were male and right-hand dominant, with ages ranging from 23 to 56 years old. They provided information regarding their previous experience in the different types of arthroscopic surgeries. In general, participants were divided into three subgroups based on their training level and experience in arthroscopy, viz., junior doctors pursuing a higher degree and residents undergoing specialty training were classified as the novice group, junior specialist surgeons were classified as the intermediate group, and consultant surgeons and senior consultant surgeons were classified as the experienced group; their relevant experience in arthroscopy of the different joints is shown in Table 1. However, two junior specialist surgeons who had significantly less arthroscopy experience than their peers were categorized into the novice group. One consultant surgeon who had no prior experience in hip arthroscopy was placed into the intermediate group. Based on power calculation for one-way ANOVA comparing three groups, a sample size of six was needed in each group to obtain a power of 0.8 when the effect size was 0.9 and a significance level of 0.05 was employed.

Before the start of the simulated operation, a 4-h didactic lecture on hip arthroscopic surgery was provided to the participants. The contents covered in the lecture included relevant anatomical and operational knowledge, which was intended to help participants familiarize themselves with the steps that they would be performing on the simulator. For the simulated operation, the demo version of an arthroscope was used, with other standard surgical equipment and tools, such as fluoroscopic equipment, to enhance the accuracy of the simulation.

An arthroscopic surgeon with 2 years of experience performing hip arthroscopy assumed the role of an instructor to guide participants in performing tasks with the simulator. A task-specific checklist (Supplemental Table 1), created by two experienced staff surgeons based on an earlier study [25], was used to guide and evaluate the operational processes. The instructor graded the performance of participants using this task-specific checklist. The time taken for task completion and the number of times fluoroscopy was used were also noted by the instructor. Hand and arthroscopic footage for each simulated operation were videotaped. Another arthroscopic surgeon with 5 years of experience in hip arthroscopy served as a blinded assessor to review the deidentified videos after the study and graded each operation using the Arthroscopic Surgical Skill Evaluation Tool (ASSET) [15] and final global rating scale (GRS) [12]. After the simulated operation, a poststudy qualitative survey using a five-point Likert scale with anchor statements was conducted to collect feedback from the participants about their perception and utility of the simulator (Supplemental Table 2).

The total scores for the task-specific checklist, ASSET, final GRS and feedback (with maximum scores of 15, 38, 5 and 20, respectively) were collated, and mean scores were calculated. The internal consistency and reliability for the task-specific checklist and ASSET scores were measured using Cronbach’s alpha. The differences in performance and feedback scores between the three groups of participants were assessed using one-way ANOVA with a post hoc Tukey test. The Pearson correlation coefficient was used to calculate the correlation between the task-specific checklist, ASSET and final GRS scores. The effect size was measured by Cohen's f statistic for one-way ANOVA. All statistical analyses were performed using R software, version 4.1.1 (R Foundation).