Osseous free tissue reconstruction for the mandible was proposed by Hidalgo and Pusic in the early 1970s as an effective method for mandibular reconstruction, and thus has become the standard of care.1 Compared to alternative options, Kroll et al. also determined vascularized bone grafts to be the preferred method for reducing costs and the postoperative complication rate.2
Mandibular reconstruction requires a particularly high level of precision to achieve optimal functional, structural, and aesthetic outcomes. Because of its complex nature, modern technologies have become popularized to further refine the procedure and improve outcomes.3 Specifically, preoperative computerized planning and computed tomographic imaging of the donor site and native mandible—or virtual surgical planning—have been used for improving precision in free flap reconstruction of mandibular and maxillary defects.3–7 Patients with mandibular discontinuity or complex oncologic disease obviating preplating of the mandible are believed to be the most appropriate for virtual surgical planning. These include children with congenital deformities, patients with facial gunshot wounds, and head and neck oncology patients,8 especially those with outer cortex involvement or for whom multiple osteotomies are required.9
The stages of virtual surgical planning include preoperative planning through an interactive conference with a project engineer, modeling, and then surgery. The preoperative teleconference allows the surgeon to plan the proposed surgical resection, if needed; perform anatomical placement of the free flap; and design custom precontoured reconstruction plates. The materials designed as a result of this virtual meeting are the stereolithographic model of the native craniofacial skeleton and custom plates and/or cutting guides. A three-dimensional computed tomographic scan of the lower extremity may also be used for preoperative fibula osteotomy planning, allowing for harvest cutting guides. The proposed benefits of these outputs, which allow for preoperative decision-making, are observed during and after surgery. Compared with usual care, virtual surgical planning has been suggested to reduce operative and ischemia time,5–7,9–12 improve accuracy of postoperative reconstructive outcomes,3–7,9 and potentially result in fewer postoperative complications.3
This analysis aims to evaluate the cost-effectiveness of virtual surgical planning versus usual care in mandibular reconstruction surgery based on the payer perspective of the costs. There are clearly higher front-end costs with use of virtual surgical planning technology from generation of models, custom plates, and/or cutting guides. However, we hypothesize that these costs are offset by reduced operative time and reduced probability of downstream complications, both acute and delayed.
MATERIALS AND METHODS
Study Design
A decision-analytic model to comparatively understand cost-effectiveness of virtual surgical planning and usual care treatments was designed based on (1) additional costs of virtual surgical planning and (2) costs attributed to probabilities of developing postoperative complications. Clinical outcomes, including mandible infection, flap salvage, and flap failure, were considered (Fig. 1). Cost data of complications were estimated by means of the Centers for Medicare & Medicaid Services Physician Fee Schedule.13 CPT codes used are provided in Table 1.
;)
Table 1.: Cost Calculations and CPT Codes
Fig. 1.: Decision analytic model. The decision tree illustrates the progression of patients and possible acute and delayed complications for patients undergoing operative treatment—virtual surgical planning or usual care—for mandibular reconstruction surgery. VSP, virtual surgical planning.
Literature review was used to determine the basis for assumptions regarding operative/ischemia time6,9–11,14 for each treatment alternative, likelihood of complications,6,10,11,14 and cost of virtual surgical planning (Table 2).10,11,13–16 Ischemia time was selected to estimate operative times because of its established importance to promoting flap survival17 and to reducing overall complications.18 Inclusion criteria for literature review consisted of articles specific for maxillomandibular reconstruction, those conducted after 2012 to address the potential learning curve in using virtual surgical planning, and those that included a control group versus virtual surgical planning.
Table 2.: Input Parameter Table and Sensitivity Values
After obtaining institutional review board approval, a retrospective chart review of patients who underwent virtual surgical planning at the University of North Carolina at Chapel Hill from January of 2012 to December of 2015 was conducted. These data were used solely to inform model structure. We created a simulated cohort of 1000 patients based on literature review. The population of our simulated cohort from the literature was composed of patients who underwent maxillomandibular reconstruction for trauma or head and neck cancer with osteocutaneous free flaps. Type of free flap was isolated to fibula flaps. Ages varied largely across the literature (range, 16 to 88 years).
Regardless of the treatment, there is a probability of experiencing infection of the reconstructed mandible after the initial procedure and the potential for experiencing other acute and delayed complications. The delayed complications as a result of the flap being compromised may lead to (1) restoration of flap (flap salvage), (2) partial flap loss, or (3) total loss of flap (flap failure). For those patients experiencing total flap loss, there is potential for reconstruction with (1) an additional osseous free tissue transfer, (2) soft-tissue free tissue transfer, or (3) regional soft-tissue reconstruction. Some patients with flap failure may not be eligible for reconstruction and therefore undergo débridement only with no further reconstruction. Total flap loss was highlighted because of the greatest associated costs attributable to likelihood of reoperation and further reconstruction.
Because there was variation in the reporting of specific complications in the literature, our model was adjusted accordingly. Although our complications of interest included infection, flap salvage, and flap loss, these were not displayed exclusively and not presented in a mutually independent manner in the literature fitting our inclusion criteria. For instance, it was not clear from Weitz et al.9 or Culié et al.10 whether those experiencing flap loss were independent of the patients who presented with infection. Therefore, our model assumed no differences in probability of failure for those with or without infection at the site of the mandible.
Another accommodation of our model was the inclusion of “no or other delayed complications,” following infection or no infection. No complications and other complications were aggregated because of the diverse complications reported across the literature that were not relevant to our analysis, such as postoperative hemorrhage or fistula formation.9,10 Sensitivity analysis allows for consideration of the spectrum of consequential costs of this branch. Finally, because none of the studies fitting our inclusion criteria included flap salvage information, this was estimated from Toto et al.11 Although this study was outside of our intended dates of inclusion, and did not include a control group, we assumed flap salvage to be experienced similarly across all patients in our model.
Model Parameters
Values for average ischemia time, base-case probability of complications, and cost estimates are listed in Table 2, along with distributions used to describe uncertainty and possible variation in estimates. Base-case estimates were calculated as weighted averages across studies, where weights were a function of a study’s sample size.
All costs are from the payer perspective and in 2016 U.S. dollars. All dollar values were adjusted to 2016 U.S. dollars using the Consumer Price Index from the U.S. Bureau of Labor Statistics. The costs incorporate the additional cost of virtual surgical planning compared with usual care, cost of mean ischemia time for virtual surgical planning and usual care, and procedural cost for respective complications. Costs for the initial procedure were calculated based on ischemia time of respective treatment alternative and per-minute cost of surgery time estimated through the University of North Carolina at Chapel Hill. These per-minute costs incorporate the hospital’s per-minute operating room fee and the anesthesia fee.
Costs of complications were estimated using appropriate CPT codes for each respective condition. Although infection of the mandible and flap salvage were defined using singular CPT codes (débridement and neck exploration), flap failure required multiple codes to be combined because of the nature of probable steps following flap failure. Although neck exploration and débridement are always performed for flap failure incidents, the remaining costs are conditional on whether or not further reconstruction is undertaken, and which type of flap is used (bar and soft tissue, bone only, or regional tissue). The costs for types of potential reconstruction following flap loss were added to the costs of flap loss (Table 1). Costs of “other or no delayed complications” considers zero cost for the base-case event, whereas sensitivity analysis incorporates the potential for exploration for infection, fistula, or postoperative hemorrhage.
All estimates in our model were treated as uncertain and, as such, appropriate distributions and distributional parameters were selected based on available data and established best practices found in the literature for sensitivity analyses to be performed. Although the probabilities of types of post–flap failure outcomes are also uncertain, these were estimated through qualitative interviews with University of North Carolina at Chapel Hill clinicians, and therefore no sensitivity parameters could be reasonably approximated, and this branch was excluded from sensitivity analyses.
Base-Case Analysis
In the base-case analysis, point estimates of costs and probabilities of clinical complications, and costs of ischemia times (Table 2) were used to estimate the total costs and number of complications for our hypothetical cohorts progressing through each alternative. Total costs per person were calculated conditional on complication probabilities. Our primary complication of interest was flap loss, as this outcome has the greatest cost repercussions. Finally, a difference of total costs of virtual surgical planning compared to usual care was also estimated.
Threshold Analyses
Threshold analysis was performed to determine how costs and likelihood of complications for virtual surgical planning would have to change to net cost differences between usual care and virtual surgical planning. Goal Seek and Solver tools with Microsoft Excel 2010 (Microsoft Corp., Redmond, Wash.) were used for these scenario analyses. In the first scenario, we aimed to determine a threshold by which costs of virtual surgical planning would need to change to net cost differences between usual care and virtual surgical planning. In the second scenario analysis, thresholds were identified whereby complication probabilities of infection and total flap would have to be reduced to minimize total costs of virtual surgical planning.
Univariate Sensitivity Analysis
A tornado chart diagram was generated with Oracle Crystal Ball 2014 (Oracle Corp., Redwood Shores, Calif.) software to identify the parameters in the base-case analysis that most impacted the cost differences between usual care and virtual surgical planning (Fig. 2). The chart illustrates a series of one-way sensitivity analyses represented by horizontal bars; the most influential parameter in determining cost differences is depicted by the widest bar. All assumptions in the model are considered uncertain; however, those included in this analysis were limited to parameters for which distributional parameters could be estimated. These included ischemia time, all cost data, and probability estimates for complications.
Fig. 2.: Tornado diagram of univariate sensitivity analysis. Bar labels show the difference from the base case for the three most influential variables. VSP, virtual surgical planning.
The distributions for our data were selected based on available information. Assumptions of the distributions were made dependent on what is acceptable for each respective type of data. Distributions for ischemia times were selected after random sampling and bootstrapping of available University of North Carolina clinical data of virtual surgical planning procedures. We assumed that usual care procedures reflect similar distributions of ischemia time observed by the virtual surgical planning simulated data (gamma distribution). This decision was validated by best practices of analogous types of count data.14 The distributions for probabilities of complication and cost data were also selected based on best practices.14 The distributional parameters were estimated when variation in estimates was observed, and given as a standard deviation when multiple estimates of a given parameter were available. All probabilities were considered to be beta distributions and their parameters calculated based on study sample (n), number of successes (α), and β (n − α). The values for α and β in the beta distribution were calculated as follows: α = r, β = n – r.
Probabilistic Sensitivity Analysis
Using Oracle Crystal Ball 2014 software, Monte Carlo simulations were performed (1000 trials) with our 1000-patient hypothetical cohort to consider the uncertainty of varying all parameters simultaneously. The distributions and distributional parameters were unchanged from the univariate sensitivity analysis.
RESULTS
Base-Case Results
Results of base-case analysis indicated that virtual surgical planning was costlier than usual care by a difference of $7099 per person (Table 3). In addition, the virtual surgical planning group did not experience less risk of complications or flap loss. Virtual surgical planning cases had an increased incidence of flap loss (77 in the usual care group versus 83 flaps lost in the virtual surgical planning group) by 0.6 percent and an increased incidence of mandibular infection (243 for usual care versus 308 for virtual surgical planning) by 6.5 percent (Fig. 3).
Table 3.: Base-Case and Probabilistic Sensitivity Results
Fig. 3.: Detailed decision analytic model. This is the same decision tree demonstrating the costs that were calculated and the probabilities for complications that were calculated by taking weighted averages from each article, weighted by sample size. Based on all of this, the total cost difference was calculated between groups. We assumed flap salvage to be experienced similarly across all patients in our model. VSP, virtual surgical planning.
Threshold Analysis Results
An initial threshold analysis was run to understand how much virtual surgical planning costs would need to be reduced to match total costs of usual care, if complications probabilities and associated costs remain the same. This analysis revealed that the cost of virtual surgical planning material costs would have to be reduced to $712. A second threshold analysis was conducted by preserving base-case virtual surgical planning costs and adjusting probabilities of complications to understand how virtual surgical planning would optimally need to minimize probabilities of postoperative flap loss and mandibular infections to be cost-saving compared with usual care. Scenario analysis results demonstrated that no feasible solution is possible for cost differences between treatments to be canceled out. Furthermore, if both probability of total flap failure and likelihood of infection are reduced to 0 percent, there is still an excess of $6682 remaining for costs of virtual surgical planning.
Sensitivity Results
Univariate sensitivity analysis validated the impact of costs of virtual surgical planning materials toward cost difference between alternatives, revealing that costs of virtual surgical planning contribute to 82 percent of variation in cost differences (Fig. 2). Furthermore, the probabilistic sensitivity analysis generated spread of possible results after 1000 simulations (Table 3). Results of trials suggested that, given uncertainty of all parameters, virtual surgical planning is never less costly than usual care. This supported initial base-case findings. However, in 49 percent of iterations, there are fewer cases of flap loss, the costliest of complications.
DISCUSSION
Virtual surgical planning is becoming increasingly popular for use in preoperative planning for complex maxillomandibular reconstruction, primarily for its ability to streamline the intraoperative process for these particularly challenging cases. The question remains, however, of whether the potential cost-savings in operative time reduction are balanced by the cost of the planning. Virtual surgical planning itself has a highly variable cost, with reports of $800 to $8200 per case.11,12,16,19 This upfront cost depends on what the surgeon designs and orders during the planning session, such as the number of cutting guides and models, and the prebent plate specifications such as thickness and length, and is also affected based on the modeling company that is used. Although we found that virtual surgical planning was costlier than usual care by a difference of $7099, it is important to keep in mind that the overall hospital costs associated with these types of reconstructions are reported to range anywhere from over $20,00011 to $50,000.2 In the grand scheme, the extra cost of virtual surgical planning is only a fraction of the total cost of care these patients required and, at least for larger institutions, this extra cost is usually bundled into the price of the procedure.
Virtual surgical planning has been shown to reduce operative time by as much as 2.5 hours,11 but the relative cost savings of this reduction in operating room time alone may not offset the average cost of virtual surgical planning. Several studies have demonstrated shorter overall operative times with virtual surgical planning use, both total operative time6,10,11 and mean time from anastomosis to the end of the procedure9 (Table 4). Although ischemia time was highlighted in this study because of its established importance in promoting flap survival17 and in reducing overall complication rate,18 total surgical time is important to note, as increased overall duration of operating time has been associated with increased rates of surgical-site infections and thromboembolic events.
Table 4.: Operative Time for the Virtual Surgical Planning Group versus the Conventional Group
Although previous literature has proposed that the benefits of preoperative decision-making that virtual surgical planning allows may reduce long-term costs because of improved reconstruction accuracy and reduced free flap ischemia time,8 this conclusion was not supported in our analysis. We found virtual surgical planning to be costlier and approximately equal to usual care in reducing the outcome of total flap loss. Total flap loss was specifically highlighted because of the greatest associated costs attributable to the likelihood of reoperation and reconstruction. From the patient perspective, the difference in cost of $6490 with virtual surgical planning from our probabilistic sensitivity results (Table 3) may be worth reducing the likelihood of burdensome complications and further hospitalizations in the long term, but our data did not support this inference. In fact, our data showed an increased incidence of flap loss and mandibular infection in the virtual surgical planning cohort; this may be attributable to a number of issues. For one, virtual surgical planning patients often require a more complex reconstruction.8,9 Three of the four studies used for our input parameters had patients of similar complexity undergoing reconstruction with virtual surgical planning and conventional methods.6,10,11 Weitz et al. had overall similar patients in each cohort, except that the virtual surgical planning group did have a significantly greater number of patients requiring two or more osteotomies versus the conventional group.9
Because virtual surgical planning technology is fairly novel and a continually evolving planning tool in mandibular reconstruction,4 there are inherent limitations in our study. To address the potential learning curve in using virtual surgical planning from the clinical perspective and associated patient outcomes, we restricted our literature review inclusion criteria to the most recent studies, after 2012, to parallel available University of North Carolina at Chapel Hill clinical data (except for flap salvage probability, wherein no current literature was available and therefore it was assumed that there was no difference between treatments). Still, virtual surgical planning is experiencing continual refinement, and surgeon experience can certainly impact length of operating room time and flap failure and infection rates. Further research needs to be undertaken to better reflect current outcomes of virtual surgical planning use.
Our lack of University of North Carolina at Chapel Hill control data may have contributed to sampling bias in identifying distribution because more complex patients, as defined by institution, usually receive virtual surgical planning over usual care. We were unable to control for clinical and institutional variation that may exist in how virtual surgical planning is selected and used in our literature review. Although several studies have suggested indications for use of virtual surgical planning,8–10 set guidelines for these complex reconstruction patients have not been established regarding virtual surgical planning use and therefore surgeon preference and comfort level will continue to guide choice as to whether virtual surgical planning is used or not. Additional refinement of patient complexity definitions and delineation between less and more complex patients would assist in establishing best practices for decision-makers.
In addition to lack of clinical control data, another limitation was the general lack of consistent data across the literature, which is likely attributable to the novelty of virtual surgical planning. Although we limited study dates, the lack of data on this topic required our inclusion criteria to be broad, such that we were not able to limit for clinical diagnoses, as would have been ideal for controlling any potential confounding variables. A larger sample size from one institutional randomized controlled trial would have been optimal to also control for institutional and clinical differences.
Uncertainty in the model structure was also a limitation in this study. Because of variation in reporting complications across the literature, there were challenges in model construction related to accounting for postoperative complications. Lastly, there was uncertainty in the distributional assumptions made which, although validated through bootstrapping of available data, were biased in that they consisted only of treatment information.
CONCLUSIONS
This cost-effectiveness analysis brings attention to the expenses related to virtual surgical planning, which do not currently seem to correlate with patient outcomes in reducing downstream complications. This has been demonstrated in several articles; despite more accurate bony contact on postoperative imaging with virtual surgical planning,3–7,9 overall functional and aesthetic results do not differ between groups.5,6,10,11 Although some plastic surgeons “feel that no price tag can be assigned to predictability and streamlining”8 regarding the value of virtual surgical planning in complex mandibular reconstruction, results warrant further investigation to determine which patient population is best served by this technology, especially given today’s escalating health care costs. To minimize complications, virtual surgical planning should be used as an adjunct to and not as a substitute for surgical experience and expertise in complex mandibular reconstruction.
REFERENCES
1. Hidalgo DA, Pusic AL. Free-flap mandibular reconstruction: A 10-year follow-up study. Plast Reconstr Surg. 2002;110:438–449; discussion 450451.
2. Kroll SS, Schusterman MA, Reece GP. Costs and complications in mandibular reconstruction. Ann Plast Surg. 1992;29:341–347.
3. Roser SM, Ramachandra S, Blair H, et al. The accuracy of virtual surgical planning in free fibula mandibular reconstruction: Comparison of planned and final results. J Oral Maxillofac Surg. 2010;68:2824–2832.
4. Hirsch DL, Garfein ES, Christensen AM, Weimer KA, Saddeh PB, Levine JP. Use of computer-aided design and computer-aided manufacturing to produce orthognathically ideal surgical outcomes: A paradigm shift in head and neck reconstruction. J Oral Maxillofac Surg. 2009;67:2115–2122.
5. Chang EI, Jenkins MP, Patel SA, Topham NS. Long-term operative outcomes of preoperative computed tomography-guided virtual surgical planning for osteocutaneous free flap mandible reconstruction. Plast Reconstr Surg. 2016;137:619–623.
6. Wang YY, Fan S, Zhang HQ, Lin ZY, Ye JT, Li JS. Virtual surgical planning in precise maxillary reconstruction with vascularized fibular graft after tumor ablation. J Oral Maxillofac Surg. 2016;74:1255–1264.
7. Zhang L, Liu Z, Li B, Yu H, Shen SG, Wang X. Evaluation of computer-assisted mandibular reconstruction with vascularized fibular flap compared to conventional surgery. Oral Surg Oral Med Oral Pathol Oral Radiol. 2016;121:139–148.
8. Saad A, Winters R, Wise MW, Dupin CL, St Hilaire H. Virtual surgical planning in complex composite maxillofacial reconstruction. Plast Reconstr Surg. 2013;132:626–633.
9. Weitz J, Bauer FJ, Hapfelmeier A, Rohleder NH, Wolff KD, Kesting MR. Accuracy of mandibular reconstruction by three-dimensional guided vascularised fibular free flap after segmental mandibulectomy. Br J Oral Maxillofac Surg. 2016;54:506–510.
10. Culié D, Dassonville O, Poissonnet G, Riss JC, Fernandez J, Bozec A. Virtual planning and guided surgery in fibular free-flap mandibular reconstruction: A 29-case series. Eur Ann Otorhinolaryngol Head Neck Dis. 2016;133:175–178.
11. Toto JM, Chang EI, Agag R, Devarajan K, Patel SA, Topham NS. Improved operative efficiency of free fibula flap mandible reconstruction with patient-specific, computer-guided preoperative planning. Head Neck 2015;37:1660–1664.
12. Honeybrook A, Sexton K, Patel A, et al. Fibula free flaps for mandibular reconstruction: A 10-year experience including the introduction of virtual surgical planning. J Maxillofac Trauma 2013;2:1–13.
13. Centers for Medicare & Medicaid Services. Physician fee schedule search. Available at:
https://www.cms.gov/apps/physician-fee-schedule/search/search-criteria.aspx. Accessed March 7, 2019.
14. Briggs A, Claxton K, Sculpher M. Decision Modelling for Health Economic Evaluation. 2006.New York: Oxford University Press.
15. Macario A. What does one minute of operating room time cost? J Clin Anesth. 2010;22:233–236.
16. Zweifel DF, Simon C, Hoarau R, Pasche P, Broome M. Are virtual planning and guided surgery for head and neck reconstruction economically viable? J Oral Maxillofac Surg. 2015;73:170–175.
17. Valentini V, Agrillo A, Battisti A, Gennaro P, Calabrese L, Iannetti G. Surgical planning in reconstruction of mandibular defect with fibula free flap: 15 patients. J Craniofac Surg. 2005;16:601–607.
18. Chang SY, Huang JJ, Tsao CK, et al. Does ischemia time affect the outcome of free fibula flaps for head and neck reconstruction? A review of 116 cases. Plast Reconstr Surg. 2010;126:1988–1995.
19. Hammoudeh JA, Howell LK, Boutros S, Scott MA, Urata MM. Current status of surgical planning for orthognathic surgery: Traditional methods versus 3D surgical planning. Plast Reconstr Surg Glob Open 2015;3:e307.
