Tumor resection within the pelvis: Accuracy study of the conventional surgical technique

INTRODUCTION Treating malignant tumors within the pelvis is challenging due to the complex anatomy of the pelvis. Nowadays, one of the most satisfactory solutions is a large resection combined with reconstruction by massive allograft. Computer-assisted surgery was introduced to increase the accuracy of orthopaedic surgical procedures. Commercially available navigation systems using preoperative computed tomography (CT) data can be used for sacroiliac screw insertion [1], pelvic tumor resection [2], pelvic osteotomies [3] or pelvic ring fracture reduction[4]. Navigation system increases accuracy as it was demonstrated in pedicle screw insertion compared to conventional surgical technique [5]. No experimental data exist about the accuracy of pelvic tumor resection using conventional technique. This lack of data renders difficult performance evaluation of computer and robotic assisted technologies. This experimental study tends to report the accuracy of tumor resection within the pelvis and reconstruction by massive allograft with conventional surgical technique. An experimental model on sawbones was used to simulate tumor resection and new methods were created to evaluate the safe margin and the reconstructed host-graft junction. METHODS Four experienced surgeons (Surg1, Surg2, Surg3, Surg4) were each asked to perform resection of 3 different pelvic tumors and their reconstruction. Twelve identical pelvic sawbones and 12 left hemipelvic sawbones were considered as host bones (Host) and allografts (Graft) respectively. The 12 Hosts were scanned. Three sets of tumor were virtually created in zone I, zone II and zone II-III of Enneking respectively. The tumor was represented by a sphere and placed on the 3D CT-scan of the pelvis using the visualization software VolView. The 3D CT-scan of the virtual tumor was cut in sagittal, coronal and transversal slices. A graduated scale was available on each slice. The surgeon could plan and perform resection of the 3 tumors without limitation in time. Instruction was given to respect as accurately as possible a 10mm-safe margin from the tumor. Based on the 2D slices, he could note some landmarks with a skin marker on the sawbone to guide the cutting. After planning and landmarking, he had to perform tumor resection by cutting the sawbone using an oscillating saw. After tumor resection, the resected parts were scanned and registered with the 3D CT-scan of the corresponding Host. The minimal distance between tumor and resection planes was measured with Volview. This allowed to verify if the 10mm-safe margin had been respected or not. A negative value was given for intralesional cutting and positive for extralesional cutting. Each surgeon had to reconstruct the 3 pelvis. Oversized Grafts were given to increase technical difficulty. The surgeon could note some landmarks to guide the cutting of the graft. Three additional cutting were allowed to each surgeon to optimize the reconstruction. The Host-Graft reconstruction was temporarily osteosynthetized with 2mm K-wires. Each Host-Graft junction (HGJ) was evaluated by 2 different methods. Firstly, the maximal gap between Host and Graft was measured for all the HGJ with an electronic caliper by 3 different observers (Obs1, Obs2, Obs3). Secondly, an epoxy paste of high density was used to fill the gap of the HGJ. All the constructs were scanned. As the density of the paste was very different from the sawbone, a threshold segmentation eliminated the voxels corresponding to the sawbone. The number of voxels corresponding to the paste were calculated, giving the gap volume between Host and Graft. To compare the measures of the maximal gap by the 3 observers, paired Student t-test were realized for each pair of observers. To compare the results of the 4 surgeons (safe margin, maximal gap, gap volume), paired Student t-test were performed for each pair of surgeons. The chosen level of significance was 0.05. RESULTS Twenty-four HGJ were available for evaluation : 1 HGJ for zone I-tumor, 2 for zone II-III-tumor and 3 for zone II-tumor. All but 2 resections (of the 12) were “carcinological” with respect of a 10mm-safe margin. Two resections of zone I-tumor were “intralesional” (safe margin of -5.02mm and -5.14mm respectively). The mean safe margins for each surgeon are summarized in Table I. The differences between surgeons are not significant. The mean maximal gap between Host and Graft for the 4 surgeons are summarized in table I. There are no significant differences between surgeons and between observers. Detailed data about the mean gap volume of the HGJ for the 4 surgeons are summarized in Table I. These differences are not significant. DISCUSSION New methods were designed to calculate safe margin of tumor resection, maximal gap and gap volume of the Host-Graft junction. The results show the lack of accuracy during resection. This can be explained by the difficulty for the surgeon to transfer planning from 2D slices to a complex 3D pelvis. It is also hard for the surgeon to reproduce the same cutting planes on the graft. Generally, the planes were not perfectly parallel, giving a gap between host and graft. No surgeon had significantly better results than the others. Two surgeons performed an intralesional resection, demonstrating the lack of accuracy. These data could be considered as reference database for further evaluation of computer and robotic assisted tumor resection within the pelvis