Navigation Guidance During Free Flap Mandibular Reconstruction: A Cadaveric Trial

Key Points

Question  Is navigation-assisted reconstruction accuracy similar to template-assisted but better than freehand reconstruction of mandible defects resulting from resection of tumors effacing the buccal cortex?

Findings  In this trial that included 10 cadavers, navigation-assisted ramus realignment was within 0.54 mm of template-assisted reconstruction. Two of 4 cephalometric distance measurements were likely improved for navigation-assisted compared with freehand technique.

Meaning  Navigation-assisted mandible reconstruction offers a reliable technique for real-time, computed tomography–based alignment when reconstructing oncologic defects resulting from tumors effacing the buccal cortex.

Importance  Segmental mandibulectomy for tumors that distort the buccal surface of the mandible present a reconstructive challenge.

Objective  To determine whether mandible alignment after navigation-guided mandible reconstruction is better than alignment after non–template-assisted freehand reconstruction and as good as template-assisted reconstruction in a cadaveric trial.

Design, Setting, and Participants  A cadaveric trial using 10 specimens was conducted at a tertiary academic center. Fiducials were created on the ramus to compare alignment with each intervention. Segmental mandibulectomy was performed on each cadaver. Each cadaver underwent navigation-guided reconstruction, template-assisted reconstruction using a manually shaped plate, and non–template-assisted freehand reconstruction with plate contouring performed after mandibulectomy. The study was conducted from October 1, 2015, to January 1, 2016; data analysis was performed from February 1, 2016, to March 1, 2016.

Interventions  Segmental mandibulectomy, navigation-guided reconstruction, template-assisted reconstruction using a manually shaped plate, and non–template-assisted freehand reconstruction.

Main Outcomes and Measures  Ramus fiducial coordinates were recorded at baseline and after each intervention. Mandible dimensions were measured using cephalometric landmarks. Postintervention and baseline differences in ramus and mandible position were calculated.

Results  Ramus alignment was not significantly different between navigation-guided and template-assisted reconstruction, differing by 0.54 mm (98.3% CI, −0.38 to 1.47 mm). Non–template-assisted freehand reconstruction was associated with a 3.14-mm difference in alignment compared with template-assisted reconstruction (98.3% CI, 1.09 to 5.19 mm). Navigation-guided alignment resulted in a 3.69-mm improvement in alignment compared with non–template-assisted freehand reconstruction (98.3% CI, 1.79 to 5.58 mm). There was some improvement in the gonion-gonion and lingula mandibulae–lingula mandibulae (Lm-Lm) alignment for navigation-assisted compared with non–template-assisted freehand reconstruction by 1.97 mm (98.3% CI, −0.65 to 4.58 mm) and 1.39 mm (98.3% CI, −0.17 to 2.95 mm), respectively. There was marginal evidence of better Lm-Lm alignment for navigation-guided than template-assisted reconstruction (0.44 mm; 98.3% CI, −0.06 to 0.95 mm).

Conclusions and Relevance  Mandible alignment following navigation-guided reconstruction is similar to template-assisted reconstruction. Navigation-guided alignment is likely better than non–template-assisted freehand reconstruction, and navigation guidance offers a reliable technique for real-time adjustment when reconstructing complex surgical defects, such as tumors effacing the buccal cortex of the mandible.

Template-assisted plate bending for segmental mandible reconstruction is precluded when distortion of the native mandible architecture has occurred.^1,2 Lacking the native contour of the mandible to guide plate bending, alternative techniques are needed to re-establish precise mandibular alignment and occlusion. Recently, investigators^1,3,4 evaluated the efficacy of computer-aided planning and navigation guidance for cases of buccal cortex effacement and noted improved reconstruction accuracy; however, controlling for confounding is challenging in case series. Advantages of navigation-guided mandible reconstruction include the accessibility of navigation systems, decreased need for engineering support, decreased costs associated with preprinted models and templates, and a short setup time.

Current techniques for free flap reconstruction after resection of tumors effacing the buccal cortex are limited by the inability to adjust the reconstruction plan intraoperatively. Temporary external fixation relies on adequate bone stock to position the fixator and reconstruction plate.^5,6 Inadequate bone stock may interfere with placement of a reconstruction plate, and segmental mandibulectomy incorporating the ramus inhibits placement of an external fixator near or on the condyle. Computer-assisted design (CAD) with 3-dimensional (3D) models and prebent reconstruction plates allow for accurate reconstruction via computer-optimized virtual and physical models.^7-10 However, CAD requires added time for preoperative engineering consultations and adds costs in excess of $10 000, and models may not be adjusted in the case of incorrect positioning of cutting guides¹¹ or updated resection margins. Moreover, there are several steps involved in the process of transferring information from the virtual to physical model creation and implementation in free flap mandible reconstruction, with each step harboring potential for error. In contrast, navigation-assisted reconstruction implements computed tomographic (CT) guidance, thus allowing for real-time, patient-specific adjustments to mandible position during reconstruction of defects effacing the buccal cortex.

In the present study, we sought to investigate the utility of navigation-assisted mandible reconstruction using an experimental cadaver model. A cadaver model was selected to control for factors including defect size, reconstruction type, and surgeon variability. In addition, each cadaver served as its own control, which would not be possible in a randomized clinical trial in patients. We hypothesized that mandible alignment after navigation-guided reconstruction would be as good as alignment after template-assisted reconstruction but more accurate than non–template-assisted freehand reconstruction of defects resulting from ablation of tumors effacing the buccal cortex.

Ten formalin-fixed cadaver specimens were placed in wired maxillomandibular fixation with dental splints to stabilize fiducial registration markers. Cadaver specimens underwent preoperative CT scanning while in maxillomandibular fixation. The study was conducted from October 1, 2015, to January 1, 2016; data analysis was performed from February 1, 2016, to March 1, 2016. University of Washington Institutional Review Board approval was obtained prior to initiating this study.

Three types of mandible reconstruction were performed on each cadaver: (1) template-assisted reconstruction, (2) non–template-assisted freehand reconstruction, and (3) navigation-guided reconstruction. Each cadaver underwent all 3 procedures in random order on the same hemimandible. A pseudorandom number generator was used to randomize the order of interventions within each cadaver.¹² Only 1 intervention was performed each day to mitigate surgical bias.

All procedures were performed at the University of Washington Institute for Simulation in Healthcare. The following procedures were performed on each cadaver before the test interventions. Initially, all cadavers had a reconstruction plate manually bent using the contours of the intact mandible to serve as a control modeling an optimized reconstruction. This procedure is referred to as the template-assisted technique henceforth. For the template-assisted technique, reconstruction plates that were not bent beforehand were manually bent with the aid of a malleable template, and plates were secured with 2.0-mm screws on each mandible before segmental mandibulectomy using 2.8-mm reconstruction plates. At least 2 drill holes were made on both sides of the reconstruction plate near the symphysis and along the ramus. The plates were removed and labeled according to their respective cadaver. Plates were stored for later use, and the mandible drill holes were marked with a permanent paint marker. Four fiducial points were drilled on the symphyseal and parasymphyseal regions in a square configuration. An additional 4 fiducial points were drilled near the angle and along the posterior-inferior, anterior-inferior, and anterior-superior ramus to measure ramus shift at baseline and after each intervention.

Specimens were positioned on the operating table and secured (Mayfield Triad Skull Clamp; Integra LifeSciences Corp). A tracker (Stryker Universal Tracker, Stryker Navigation System II; Stryker) was rigidly secured to the skull clamp. Preoperative CT scans were uploaded to a commercial navigation system (Stryker Navigation System II; Stryker). Cadavers were registered to the navigation system, and a pointing instrument was calibrated (Figure 1A). Registration points were visually confirmed by assessing the location of the pointer tip on the navigation system at fixed points on the cadaver, including the occlusal surface of the interincisor space, the sigmoid notch bilaterally, and the medial canthi. Euclidian coordinates were recorded at each fiducial point at baseline and before each intervention.

Registration accuracy was evaluated in a separate experiment using 1 cadaver. Euclidian coordinates of 3 fiducial points were recorded. The registration sequence was repeated 4 times. Euclidian coordinates of each fiducial point were recorded after each registration. The specimen was removed from the skull clamp and replaced between registrations, and the navigation computer was restarted. Fiducial coordinates were compared between registrations, and the position difference was calculated using MATLAB, version R2015A (MathWorks Inc). The median error was 1.8 mm (interquartile range, 0.73-2.33 mm). This value is comparable to previously reported navigation errors.¹³

Each cadaver underwent a unilateral segmental mandibulectomy alternating between a small or large defect. Small defects extended from the ascending ramus to the mental foramen. Large defects extended from the ascending ramus to the symphysis. The muscular insertions were released from the mandible to fully mobilize the ramus. After each intervention, cadavers underwent further CT scanning for analysis of mandibular dimensions relative to baseline. Template-assisted reconstruction involved placing the previously bent reconstruction plate on the mandible and securing the plate using previously drilled and marked holes (Figure 1B). Non–template-assisted freehand reconstruction involved manually bending a 2.8-mm reconstruction plate after segmental mandibulectomy, without the aid of a template, under the direction of an experienced microvascular surgeon (J.J.H.). This plate was secured using 2.0-mm screws. For navigation-guided reconstruction, we used virtual annotation markers to align the ramus with its premandibulectomy position. The annotation markers were the coordinates of the fiducial wells that were stored in the navigation computer as illustrated in Figure 2. A navigation probe was placed in the fiducial wells on the ramus, and navigation guidance was used to correlate the probe tip with the position of the annotation marker created in the baseline condition. Figure 2D illustrates this concept; the blue probe tip is aligned with coordinate 1, which corresponds to the first annotation marker created for that specimen on the ramus in the baseline condition. Synthetic fibula segments were fashioned to fill the defect, and 2.8-mm miniplates were used to secure the fibula segments (Figure 1C). The final position of the ramus and fibula segments were adjusted using the navigation probe to restore ramus alignment relative to the baseline annotation marker coordinates displayed on the navigation monitor. The screw holes marked for placement of the standard reconstruction plate were not used for the freehand plate. There was sufficient space left for the freehand reconstruction plate, allowing placement of at least 2 screws. Similarly, we were able to place at least 2 screws on each side of bone to secure miniplates during navigation-guided reconstruction.

Fiducial coordinates were recorded at baseline prior to mandibulectomy and after each intervention. Coordinates were imported from the navigation computer into MATLAB. Computed tomographic scans were imported into 3D Slicer, version 4.4.0, and 3D CTs were created.¹⁴ The 3D CT images were used to make mandible measurements, including gonion-gonion (Go-Go), gonion-menton (Go-Me), lateral condyle–lateral condyle (Lco-Lco), and lingula mandibulae–lingula mandibulae (Lm-Lm) distances.¹⁴ Fiducial markers were created on mandible landmarks of the 3D image to calculate mandible dimensions. Landmarks were defined as described previously.¹⁵ Each measurement was repeated 3 times and the mean distance was calculated.

Ramus alignment was calculated using the baseline fiducial coordinates as the reference. The distance between 2 points in Euclidian space was calculated using the formula:

.

Ramus alignment was calculated by comparing the distance between mean fiducial coordinates at baseline and after each intervention (Figure 1D). We also compared baseline and postintervention differences in mandible dimensions based on cephalometric landmarks. Our sample size estimate was determined based on previously published data.¹⁶ Seven cadavers per group were required to provide an 80% power for detecting a 4.9-mm difference in mandible alignment at a 2-sided α = .05. Ten cadavers per group were used to increase the power of the study. Regression analyses with robust SEs were used to assess mandible shift. To account for multiple comparisons, a Bonferroni-adjusted α = .017 was used for inferential hypothesis testing, and 98.3% CIs were calculated to reflect the adjusted α value. All statistical analyses were performed using Stata, version 14 (StataCorp LP).

Ramus alignment relative to baseline is illustrated in Figure 3. The mean difference in ramus position (reference, baseline) when comparing navigation-guided with template-assisted reconstruction was 0.54 mm (98.3% CI, −0.38 to 1.47 mm). The data provide evidence that the mean difference in ramus alignment was worse after non–template-assisted freehand reconstruction than after template-assisted reconstruction (3.14 mm; 98.3% CI, 1.09 to 5.19 mm). The mean difference in ramus alignment was also better for navigation-guided than for non–template-assisted freehand reconstruction (3.69 mm; 98.3% CI, 1.79 to 5.58 mm).

In a subanalysis of these data based on defect size, we evaluated changes in ramus alignment for small (n = 5) and large (n = 5) defects. For small defects, the mean change in ramus position is illustrated in Figure 4A. The data provide insufficient evidence of a difference in ramus alignment relative to baseline for navigation-guided vs template-assisted reconstruction. The mean difference was 0.30 mm (98.3% CI, −0.61 to 1.21 mm). Non–template-assisted freehand reconstruction alignment was associated with 2.36-mm worse alignment than template-assisted, although this finding was not significant (98.3% CI, −0.86 to 5.59 mm). The mean difference in ramus alignment was slightly worse for non–template-assisted freehand compared with navigation-guided reconstruction by 2.66 mm (98.3% CI, −0.55 to 5.88 mm), although it was not statistically significant.

For large defects, the mean change in ramus alignment is shown in Figure 4B. There was little evidence of a difference in ramus alignment from baseline for navigation-guided compared with template-assisted reconstruction (0.79 mm; 98.3% CI, −1.08 to 2.65 mm). In contrast, non–template-assisted freehand was worse than template-assisted reconstruction by 3.92 mm (98.3% CI, 1.13 to 6.71 mm). There was strong evidence of improved ramus alignment for navigation-guided compared with non–template-assisted freehand reconstruction (4.71 mm; 98.3% CI, 2.49 to 6.92 mm).

Given the potential for confounding introduced by varying defect size, we performed an additional analysis evaluating the overall variance contributed by defect size. We observed that the defect size accounted for 3.7% of the variance in our overall sample. In addition, we included defect size as a covariate in a regression model evaluating the association between intervention type and ramus alignment and noted that there was no strong evidence of a difference in ramus alignment attributable to defect size. The mean difference in ramus alignment between small and large defects was 0.83 mm (98.3% CI, −0.52 to 2.18 mm) when controlling for intervention.

To evaluate alignment of the contralateral hemimandible, we recorded fiducial coordinates on the symphysis at baseline and after each intervention and measured the mean coordinate alignment change from baseline. The mean (SD) differences were 2.43 (1.25) mm, 2.28 (1.61) mm, and 2.16 (1.41) mm for standard, navigation-guided, and freehand reconstruction, respectively. There was a lack of evidence to suggest a difference between any 2 groups with 1-way analysis of variance testing. Given the consistency of contralateral hemimandible alignment for each type of procedure, we did not normalize mandible alignment to alignment of the contralateral hemimandible.

To assess the association between reconstruction method and mandible dimensions, we evaluated the change in cephalometric distances. Mean differences are presented in the Table. We observed mean (SD) changes in mandible measurements relative to baseline in the transverse dimension (ie, Go-Go) up to 3.07 (3.02) mm for the non–template-assisted freehand technique compared with 1.11 (0.66) and 1.46 (1.36) mm for the navigation-guided and template-assisted techniques, respectively. In the axial dimension, we observed changes in Go-Me distance of 3.23 (2.65) mm for the non–template-assisted freehand technique compared with 2.00 (1.42) mm for the navigation-guided technique and 3.46 (2.92) mm for the template-assisted technique. There was insufficient evidence to suggest a difference in Go-Me alignment for navigation-guided reconstruction compared with template-assisted reconstruction (1.46 mm; 98.3% CI, −1.16 to 4.08 mm) or for non–template-assisted freehand compared with template-assisted reconstruction (0.23 mm; 98.3% CI, −3.03 to 3.48 mm). Similarly, the data did not suggest a difference in Go-Me change between non–template-assisted freehand and navigation-guided reconstruction (1.23 mm; 98.3% CI, −1.29 to 3.75 mm).

In the transverse dimension, change in Go-Go alignment relative to baseline was not different when comparing navigation with template-assisted reconstruction (mean change, 0.35 mm; 98.3% CI, −0.87 to 1.57 mm) or when comparing non–template-assisted freehand with template-assisted reconstruction (1.62 mm; 98.3% CI, −1.17 to 4.40 mm). There was some improvement in Go-Go alignment with navigation-guided compared with non–template-assisted reconstruction (1.97 mm; 98.3% CI, −0.65 to 4.58 mm); however, the change was not statistically significant. There was also evidence of improved Lm-Lm alignment for navigation-guided compared with template-assisted reconstruction (0.44 mm; 98.3% CI, −0.06 to 0.95 mm) and navigation-guided compared with non–template-assisted freehand reconstruction (1.39 mm; 98.3% CI, −0.17 to 2.95 mm), but again, the change was not significant. Alignment for Lm-Lm was not different between non–template-assisted freehand and template-assisted reconstruction (0.95 mm; 98.3% CI, −0.68 to 2.57 mm). Alignment for Lco-Lco was not different between the navigation-guided and template-assisted techniques (1.06 mm; 98.3% CI, −0.74 to 2.85 mm) or between navigation-guided and non–template-assisted reconstruction (0.99 mm; 98.3% CI, −2.83 to 4.81 mm). Alignment for Lco-Lco was also not different between non–template-assisted and template-assisted reconstruction (2.05 mm; 98.3% CI, −1.36 to 5.45 mm).

Our aim was to perform the initial evaluation of navigation-assisted free flap mandible reconstruction and compare it with template-assisted methods as a criterion standard. A cadaver model was chosen for 2 main reasons: (1) to minimize bias from confounding variables, such as defect size, patient factors, tumor location, and tumor size; and (2) because the performance of multiple interventions on the same patient would not be appropriate for human study. Our primary findings were that ramus alignment and mandible dimensions after navigation-assisted reconstruction were not different from those for template-assisted reconstruction but were better than for the non–template-assisted freehand technique. We observed that ramus alignment was not different when comparing navigation-guided with template-assisted reconstruction, with a difference of 0.54 mm (98.3% CI, −0.38 to 1.47 mm). We found similar observations when comparing small and large defects. We also observed up to a 3.14-mm improvement in ramus alignment and up to a 1.96-mm improvement in mandible alignment for the navigation-guided vs the freehand technique in comparison with a mean condyle-gonion alignment close to 7.3 mm that has been achieved with CAD combined with navigation-assisted techniques.¹¹

In the present study, template-assisted plating was used in lieu of external fixation because both methods would likely result in similarly excellent proximal and distal fixation. We observed reliable mandible alignment using the template-assisted technique. The mean (SD) difference in ramus alignment relative to baseline for the template-assisted technique was 2.28 (1.06) mm. To our knowledge, there have been no other studies evaluating ramus shift for direct comparison. Our findings are comparable to the results of Hanasono and Skoracki,¹⁷ who measured the position of several mandibular landmarks after CAD reconstruction and found that condyle position after CAD reconstruction was a mean (SD) 3.41 (2.86) mm different after reconstruction. In the investigators’ control group, the optimal postreconstruction alignment was for the condyle position, which differed by 3.84 (2.69) mm before vs after reconstruction. We observed increased variation in the large-defect group that was likely due to an outlier, which is discussed further below.

We noted substantial variation in our non–template-assisted freehand technique that involved manual plate bending after mandibulectomy. The mean difference in ramus alignment compared with baseline in our non–template-assisted freehand group was 3.14 mm (98.3% CI, 1.09-5.19 mm). Comparatively, in a study evaluating the utility of CAD and navigation-guided mandibular reconstruction, Yu and colleagues¹¹ noted considerable variance in their control group (ie, reconstruction based on surgeon experience) with a mean (SD) of 17.4 (3.1) mm and 12.8 (3.8) mm for condyle and gonion shift, respectively. Similarly, Hanasono and Skoracki¹⁷ found a difference in the position of bony mandibular landmarks of 6.92 (5.64) mm in their control group. This wide intersurgeon and intrasurgeon variance in mandible alignment after reconstruction may be related to the lack of a reference from native mandible contours, surgical exposure, challenges with manual plate bending, experience, and/or the type of tissue used for reconstruction. An advantage of the technique that we have presented is that it decreases the variance in reconstruction accuracy by utilizing navigation equipment available at most tertiary care institutions.

We also observed a wide variation in ramus alignment with the non–template-assisted freehand technique between large and small defects, with a larger variance for the small defects. The difference is likely accounted for by random variation that may occur even in small defects, as we observed. Manual plate bending without the contours of the intact mandible acting as a scaffold is prone to error owing to the lack of a physical template for plate shaping. One of the strengths of the navigation-guided technique is that it provides a virtual template using the patient’s CT scan and decreases the random variability associated with non–template-assisted freehand manual plate shaping.

Clinical implications of the present report include increased evidence of the use of navigation to optimize reconstructive accuracy during segmental mandible reconstruction of defects that distort or destroy contours of the native mandible. Testing the accuracy of navigation guidance compared with template-assisted techniques is critical in establishing the efficacy of navigation-guided reconstruction and providing treatment recommendations and alternative techniques to improve reconstruction. Navigation systems are found at most tertiary care centers, require a short setup time, and do not require extensive planning sessions with off-site engineers. Moreover, navigation may be used independently or complement CAD and rapid prototype manufacturing to verify placement of reconstructive free flaps and positioning of the remaining mandible and does not require proximal mandibular bone stock for maintaining alignment as does external fixation. In comparison, external fixation adds time and limitations in reconstructing defects with minimal mandibular bone stock and does not permit CT-based mandible alignment. CAD methods also improve reconstruction accuracy but do not permit real-time adjustments. Navigation guidance provides real-time feedback and can be augmented if the oncologic defect changes intraoperatively.

Limitations of the present study include added time associated with the use of navigation. The added time is similar to navigation registration in other applications, and there are no additional hardware adaptations required. In contrast, for CAD techniques, preoperative engineering consultations and time waiting for model production limit widespread applicability. In addition, we noted a wider variance in one of the ramus alignment points in the template-assisted group, which may have been the result of transportation to the hospital CT scanner and/or challenges with plate bending despite the use of a malleable template shaped on the basis of the mandibular contours. Similarly, there was greater variation in ramus alignment for the non–template-assisted freehand technique, with a larger variance for small defects compared with large defects, as discussed above. The advantage of navigation-guided reconstruction is that real-time changes in the plate shape can be performed even after mandibulectomy. In comparison, we anticipate that using an external fixator would result in accuracy similar to the template-assisted technique and improved accuracy compared with the non–template-assisted freehand technique owing to the lack of normal mandibular contours available for plate bending. In addition, we cannot follow postoperative outcomes, limiting the generalizability of this study. Last, non–template-assisted freehand reconstruction is likely surgeon dependent. Even within a single-surgeon experience, there is a wide variance in reconstruction accuracy that is attenuated with navigation guidance.

In the present study, we noted that navigation-guided mandible reconstruction provides reconstruction accuracy similar to that of template-assisted reconstruction and may be more accurate than non–template-assisted freehand reconstruction of segmental mandibulectomy defects. Variation in mandible alignment may be decreased with the use of navigation. These data support further investigation of the use of navigation guidance with or without CAD techniques during mandible reconstruction for cases with distortion or destruction of the buccal cortex of the mandible. Future work should assess the effectiveness of navigation-guided reconstruction on functional outcomes in patients undergoing free flap reconstruction of segmental mandibulectomy defects.

Corresponding Author: Jeffrey J. Houlton, MD, Department of Otolaryngology–Head and Neck Surgery, University of Washington, 1959 NE Pacific St, PO Box 356515, Seattle, WA 98195 (jhoulton@uw.edu).

Accepted for Publication: August 23, 2016.

Correction: This article was corrected on March 16, 2017, to fix errors of data reporting in the Results section of the abstract and the Results and Discussion sections of the article.

Published Online: November 17, 2016. doi:10.1001/jamaoto.2016.3204

Author Contributions: Drs Harbison and Houlton had full access to all the data in the study and take full responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Harbison, Shan, Li, Moe, Houlton.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Harbison, Houlton.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Harbison, Douglas, Li.

Administrative, technical, or material support: Harbison, Shan, Bevans, Li, Futran.

Study supervision: Bevans, Moe, Futran, Houlton.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Moe is a founder and equity holder of SpiSurgical LLC. No other disclosures were reported.

Funding/Support: This work was supported by grant T32DC000018 from the National Institutes of Health.

Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We acknowledge the University of Washington Department of Biostatistics for their advice regarding study analysis planning, the Harborview Medical Center radiology technicians for their support in computed tomographic scanning of cadaver specimens, and the staff of the University of Washington Institute for Simulation in Healthcare for assistance with equipment maintenance and setup in the surgical simulation laboratory.