Accepted Manuscript. High-volume surgeons deliver larger surgical margins in oral cavity cancer

Transcription

1 Accepted Manuscript High-volume surgeons deliver larger surgical margins in oral cavity cancer Owen G. Ellis, Michael C. David, Daniel J. Park, Martin D. Batstone PII: S (16)00118-X DOI: /j.joms Reference: YJOMS To appear in: Journal of Oral and Maxillofacial Surgery Received Date: 16 September 2015 Revised Date: 28 December 2015 Accepted Date: 15 January 2016 Please cite this article as: Ellis OG, David MC, Park DJ, Batstone MD, High-volume surgeons deliver larger surgical margins in oral cavity cancer, Journal of Oral and Maxillofacial Surgery (2016), doi: /j.joms This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

2 High-volume surgeons deliver larger surgical margins in oral cavity cancer Authors: 1. Dr Owen G Ellis a,b, MBBS, BDSc(Hons) Registrar Oral & Maxillofacial Surgery, RBWH 2. Dr Michael C David b MEd, MEpi, MSc, PhD Biostatistician, School of Public Health, UQ 3. Dr Daniel J Park b MBBS Clinical academic staff, School of Medicine, UQ 4. Associate Professor Martin D Batstone a,b MBBS, BDSc(Hons), MPhil(Surg), FRACDS(OMS), FRCS(OMFS) Director of the Department of Oral & Maxillofacial Surgery, RBWH a: Department of Oral and Maxillofacial Surgery, Royal Brisbane and Women s Hospital, Butterfield Street, Herston 4006, Australia b: The University of Queensland, St Lucia 4072, Australia Corresponding Author: Owen Ellis Tel: Mob: address: owen.ellis@gmail.com Postal address: 3/903 Brunswick Street, New Farm, QLD 4005, Australia

3 High-volume surgeons deliver larger surgical margins in oral cavity cancer Authors: Owen G Ellis a,b, Michael C David b, Daniel J Park b, Martin D Batstone a,b a: Department of Oral and Maxillofacial Surgery, Royal Brisbane and Women s Hospital, Butterfield Street, Herston 4006, Australia b: The University of Queensland, St Lucia 4072, Australia Corresponding Author: Owen Ellis Tel: address: owen.ellis@gmail.com

4 Conflict of Interest Statement: None declared

5 ABSTRACT Purpose: Many factors are involved in oral cavity cancer surgery, and their association with the surgical margin; thus, patient outcomes are still not completely understood. This aim of this study was to determine which variable or variables had the greatest influence on increasing the size of the surgical margin. Methods: A retrospective cohort study was conducted at the Royal Brisbane and Women s Hospital of patients who underwent resective surgery for a primary oral cavity cancer between January 1, 2008 and December 31, The primary outcome variable was the surgical margin, defined as the closest distance between the surgical edge and invasive cancer. A heterogeneous set of predictor variables were identified as potentially impacting on the primary outcome variable: demographic, 5 surgical, and 7 histological variables. The data then underwent statistical analysis using univariable linear regression, and variables that were found to have a statistical association were retained in a non-interaction multivariable model. Results: 250 patients were included in this study. The results demonstrated that highvolume surgeons delivered larger surgical margins than low-volume surgeons. The single most important variable associated with larger surgical margins

6 was who performed the resective operation. The following variables were also associated with smaller surgical margins: retromolar trigone location; non- SCC cancers; perineural invasion; and a lip-split mandibulectomy surigical approach. Conclusion: There was a strong association between high-volume surgeons and larger surgical margins; supporting the rationalization of oral cavity cancer management into high-volume centers and high-volume surgeons. Keywords: Oral cancer High-volume Frozen section Surgery Free tissue flap Surgical margin

7 INTRODUCTION Oral cavity cancer is the sixth most common type of cancer in the world 1, placing a large burden on the healthcare system. Most patients are managed surgically, guided by the fundamental principle that complete tumor resection is required for successful treatment. 2 Involved margins (invasive cancer at the resection site) and close margins (invasive cancer within 5mm, but not involving the resection site) are both associated with local disease recurrence and poorer survival rates. 3, 4 Therefore, achieving clear margins (no invasive cancer within 5mm of the resection site) is highly desirable to improve patient outcomes. When patients present for treatment, many prognostic factors are already predetermined, such as tumor size, nodal involvement, distant metastases and patient age. 5 The surgical margin, however, is a significant independent prognostic factor that clinicians can influence. Therefore, understanding what alters the surgical margin is highly relevant, as we can improve patient survival rates. 4 There is very little in the current literature on this subject. Sutton et al. 4 have previously defined histological factors, which were associated with poorer surgical margins in oral cavity cancer. Perineural invasion, vascular invasion, greater tumor size, and more aggressive tumors were all found to be independently associated with close or involved surgical margins. There is

8 very little known about how specific surgical techniques and clinical practices play a role. The purpose of this study was to identify which surgical and histological factors influence the surgical margin in oral cavity cancer surgery. The authors hypothesized that variables existed that could be manipulated to improve the surgical margin. The aims of this study were to 1) measure and compare the data to see which variables correlated with a statistically larger surgical margin; 2) propose evidence-based methods to increase the surgical margin.

9 MATERIALS AND METHODS: Study Design: To address the research purpose, the investigators designed and implemented a retrospective cohort study. The study sample was derived from the population of patients who presented to the Royal Brisbane and Women s Hospital (RBWH) for the evaluation and management of oral cavity cancer between January 1, 2008 and December 31, Patients eligible for study inclusion had histologically demonstrated malignant primary oral cavity cancer, which had been surgically resected at the RBWH. Patients with primary malignant melanoma of the oral cavity were excluded from the study. The RBWH Human Research Ethics Committee granted formal approval for this research project. Study Variables: In this study the primary outcome variable was the surgical margin, defined as the closest distance between the surgical edge of the excision specimen and histological evidence of invasive cancer. This was coded as a continuous variable in millimeters. During surgical resection of oral cavity cancer, the current practice at RBWH is to aim for a macroscopic surgical margin of at least 10mm. Intraoperative frozen section was utilized variably between different surgeons, where there was concern for potential involvement of surgical margins. Post-operatively the excision specimens were fixed in formalin and analyzed by the RBWH anatomical pathology unit. Although formalin can shrink specimens, and therefore alter the surgical margins, the

10 fixative procedure for all specimens in this study was standardized, having been performed in the same laboratory. A heterogeneous set of predictor variables were identified as potentially influencing the primary outcome variable. The predictor variables were grouped into sets: demographic; surgical; and histological. The only demographic information recorded was the patient age, and coded as a continuous variable in years. The surgical predictor variables were: resective surgeon; reconstructive surgeon; reconstructive type; surgical approach and access procedure. The surgeons were de-identified and categorically coded into 3 major groups, based upon the volume of cases treated: resective surgeon group 1 consisted of surgeons who treated more than 40 patients; resective surgeon group 2 consisted of surgeons who treated between 20 to 40 patients; and resective surgeon group 3 had surgeons who treated less than 20 patients. The reconstructive surgeon was categorically coded as: same (the same surgeon performed both the resection and the reconstruction); different (a different reconstructive surgeon compared to the resective surgeon repaired the defect); or nil (there was no reconstructive surgeon required). The reconstructive type was categorically coded into: nil (primary closure or no closure); graft; locoregional flap; or free flap. The surgical approach was categorically coded as per oral; mandibulectomy; mandibulotomy; lip-split mandibulotomy; lip-split mandibulectomy; drop down; or Weber-Ferguson. The surgical approaches were coded into an additional binary category, as to

11 whether they are considered access procedures or non-access procedures. Access procedures, designated yes, included mandibulotomy; lip-split mandibulotomy; lip-split mandibulectomy; drop down; and Weber- Ferguson. Non-access procedures, designated no, included per oral and mandibulectomy only. The histological predictor variables were: cancer type; location; maximum tumor dimension; T-stage; frozen section; lymphovascular invasion; and perineural invasion. The cancer type was coded in a binary fashion into: squamous cell carcinoma (SCC); or non-scc (salivary carcinoma & sarcoma). The tumor location was categorically coded into: tongue; buccal mucosa; mandible; floor of mouth; retromolar trigone; or maxilla. The maximum tumor dimension was coded as a continuous variable in millimeters. The T-stage was categorically coded into: T1 (0-20mm); T2 (20 40 mm); T3 (< 40mm); or T4 (tumor invades adjacent structures). Frozen section, lymphovascular invasion and perineural invasion were all coded in a binary fashion as either: yes or no. Data Collection & Management: Two medical practitioners, familiar with intra-operative processes and histopathology techniques, were involved in the data collection. Eligible patients were identified from the RBWH Multi-Disciplinary Team Head and Neck Cancer Committee s database, which documents all new patients and recurrences. Data was gathered from the patient s medical file; using the

12 operative report and the histopathology report. In cases where the information was not available in the patient s medical file, information was found on the hospital s intranet databases: Auslab for pathology; and ORMIS for surgical documentation. The data was collated into a single excel spreadsheet that contained the primary outcome variable and the predictor variables. All personal information was de-identified. Data Analysis: The data then underwent statistical analysis using univariable linear regression, which determined associations between the surgical margin and the predictor variables. Variables that were found to have a statistical association (p < 0.15) were retained in a non-interaction multivariable model. Age was also forced into the model, due to potential confounding. Finally, the potential effect modification was explored for each of these of variables, by testing all two-way joint effects using the likelihood ratio test. Co-linearity in the final model was tested using variance inflation factors, and model diagnostics were checked by assessing the residual distribution. Prior to analysis, a two-tailed p-value of less than 0.05 was considered to be statistically significant. Analyses were performed using Stata Version 13.0 (StataCorp, College Station, Texas).

13 RESULTS: During the study interval, 540 patients were screened for eligibility. Some patients were not offered surgery (n=116); some did not want surgery (n=16); some opted for private treatment at another site (n=130); and some did not actually have a primary oral cavity cancer in the final histopathology report (n=38). In total, 240 patients were identified for inclusion in the study, requiring a total of 250 primary resections. Of the 240 patients included, 6 patients had 2 synchronous primaries and 2 patients had 3 synchronous primaries, which were resected at different times during the study period. All patients had a histopathological report and an operative report, from which all relevant data was extracted, and no eligible patients were excluded due to incomplete data. The final sample was composed of 250 patients with a mean age of 59.4 years (SD: 13.90). Table 1 summarizes the makeup of the patient characteristics for the cohort. The results are divided into major categories: demographic variables; 5 surgical variables; and 7 histological variables. Table 2 summarizes the results for the multivariable regression analysis of the predictor variables on the surgical margin. Those variables with a p-value 0.05 were considered statistically significant. The co-efficient describes the relative impact of each variable upon the dimensions of the surgical margin (in millimeters), where positive values represent a larger surgical margin and negative values a smaller surgical margin, compared to a designated reference variable. The confidence intervals are also described. The following

14 variables were shown to have a statistically significant association with the surgical margin: volume-group resective surgeon; cancer location retromolar trigone; non-scc cancers; perineural invasion; and surgical approach - lipsplit mandibulectomy. Table 2 demonstrates that resective surgeon group 2 s surgical margins were 1.214mm (95%CI: to ) smaller than resective surgeon group 1; and that resective surgeon group 3 s surgical margins were 1.233mm (95%CI: to ) smaller than resective surgeon group 1. Table 3 indicates the mean surgical margin for each volume group of surgeons, showing that resective surgeon group 1 had significantly larger surgical margins than the other two groups. Table 4 provides the betacoefficient values, indicating which study variable had the greatest influence on the surgical margin. Table 4 lists the statistically significant study variables in descending order of influence on the surgical margin; with resective surgeon group 2 versus resective surgeon group 1, and resective surgeon group 3 versus resective surgeon group 1 at the top of the table. This indicates that the resective surgeon volume group had the strongest association of all the predictor variables on making the surgical margin larger or smaller. Table 2 indicates that cancers that were located in the retromolar trigone were found to have surgical margins that were on average 2.095mm (95%CI: to ) smaller than other locations. Non-SCC cancers had 1.960mm

15 (95%CI: to ) smaller margins on average than SCC cancers. Tumors with perineural invasion had 0.962mm (95%CI: to ) smaller margins than those cancers without perineural invasion present. Lastly, the lip-split mandibulectomy surgical approach had smaller surgical margins by 1.280mm (95%CI: to ).

16 DISCUSSION: This study s purpose was to determine if any surgical or histological variables influenced the size of the surgical margin, defined as the closest distance between the surgical edge and invasive cancer, in oral cavity cancer. The authors hypothesized that there were specific variables, as yet unknown, that could in fact be manipulated to increase the size of the surgical margin. The aims of this study were twofold: 1) measure and compare the data to see which variables correlated with a statistically larger surgical margin; 2) propose evidence-based methods to increase the surgical margin. The results of this study confirm the hypothesis that there were specific variables, which were associated with a change in size of the surgical margin. The results of the 6 statistically significant study variables shall be discussed in detail below. Furthermore, clinically relevant study variables, which had no statistically significant effect on the surgical margin, will also be discussed. From these results, the authors have proposed evidence-based practices to increase the width of the surgical margin. Statistically Significant Variables: Volume Surgery: The results of this study indicated that surgeons who had resected higher volumes of oral cavity cancer (surgeon group 1) had larger surgical margins, compared to the lower volume groups (both resective surgeon group 2 and

17 resective surgeon group 3) and that this variable was statistically the most significant amongst all variables included in the study. Volume-outcome relationships are recognized in many types of oncology surgery. 6 High volume surgeons have been shown to have improved survival rates and outcomes in: head and neck cancer 6, prostate 7, rectal 8 and breast cancer 9 surgeries. Specifically in regards to the size of the surgical margin, high volume surgeons have been shown to produce superior results in prostate oncology surgery 10. Following a review of the existing medical literature, this volume-outcome relationship has not been previously described for oral cavity cancer surgery in relation to the surgical margin. Given these significant findings, this study supports the rationalization of oral cavity cancer management into highvolume centers and high-volume surgeons on the basis of the associated increase in surgical margins, which itself is associated with lower rates of tumor recurrence and higher rates of patient survival. Cancer Location: Retromolar trigone The results of this paper demonstrated an association between smaller surgical margins and tumors that are located in the retromolar trigone. Smaller surgical margins in the retromolar trigone could be explained by the relative difficulty associated with excisions in this location. Similar results for the retromolar location have also been found recently in a large retrospective study. 3

18 Cancer Classification: In previous studies, higher rates of involved surgical margins (20%) have been described for soft tissue sarcoma (all locations). 11 Our analysis has also indicated that non-scc of the oral cavity, including both sarcomas and salivary carcinomas, were associated with smaller surgical margins than SCC. However, consideration should be given the relatively small number of non- SCC cases included in this study (n=13). Perineural Invasion: Perineural invasion is known to be associated with local tumor recurrence and decreased survival rates. 12, 13 This study has also shown an association between smaller surgical margins and perineural invasion. This association could be explained by the fact that macroscopic inspection by the resective surgeon, will not detect microscopic extensions of the tumor mass. Surgical Approach: Lip-split mandibulectomy: The results of this study indicate that tumors resected with a lip-split mandibulectomy procedure were associated with smaller surgical margins than other surgical approaches. However, the authors would suggest caution interpreting this result given the small cohort size (n = 9). Non-Statistically Significant Variables:

19 Free Flap Procedure: Performing a free flap procedure may influence the outcome of the surgical margin in two ways: It could theoretically improve the surgical margin by allowing the surgeon to remove more tissue; or conversely, it could jeopardize margins by influencing the surgeon to retain important landmarks. The influence of this factor is not completely understood. The results of this study indicate that there was no statistical difference in the surgical margin when a free flap procedure was performed (coefficient: 0.789, 95%CI: to 1.790, p-value: 0.099) These results are similar to other studies that have analyzed the surgical margin 14, and 5-year recurrence and survival rates 15, 16 for free flap procedures. Reconstructive Surgeon: When a single surgeon undertakes both the resective and reconstructive components of the operation, there could be a potential bias to retain important landmarks. Similarly, a separate resective surgeon, not involved in the reconstruction, may be more inclined to remove as much at-risk tissue as possible. Theoretically, this separation of roles may have the potential to influence the surgical margin. However, this study did not find any statistically significant relationship between the reconstructive surgeon and the surgical margin. Given this finding, there is no evidence to support the use of separate surgeons, one for resection and another for reconstruction, as a means to improve the surgical margin.

20 Frozen Section: The results of this study indicate that there is no statistically significant improvement in the surgical margin associated with intra-operative frozen section, in oral cavity cancer surgery. Frozen sections are used intraoperatively to determine a preliminary margin, with the aim of improving the definitive surgical margin. Frozen sections do, however, add to the cost of operations, both in terms of increased service provision and materials, and also increase in the duration of surgery. Many surgeons support the use of frozen section for its theoretical benefit, despite their being conflicting results in medical literature: Two prospective studies 17, 18 and one retrospective study 19 showed no improvement in the surgical margin with the use of intraoperative frozen section in oral cavity cancer; whereas, two retrospective studies 20, 21 have demonstrated improvement in the surgical margin and decreased recurrence rates. Access Procedures: Surgical access procedures aim to increase exposure to the primary oral cavity cancer by dissecting through normal facial tissue and creating a flap. They increase patient morbidity in the hope of increasing the size of the surgical margin. When the access and non-access cohorts were compared in the multivariable analysis there was no statistical difference in the surgical margin. During pre-operative planning, surgeons are more likely to select access procedures for larger tumor masses or for tumors in locations that are technically more difficult to access. The fact that access procedures have

21 similar surgical margins to non-access procedures suggests that excision of different tumor sizes has not significantly affected the surgical margin. Tumor Size: Clearly tumor size and T-stage play an important role in determining a patient s prognosis. 5 However, in keeping with the analysis of access procedure surgical margins, this study did not find a statistically significant association between tumor size or T-stage, and the surgical margin. This is in contrast to other research papers, which indicate that higher T-stage tumors are associated with higher rates of involved surgical margins. 3 There were limitations to this study, including its retrospective nature, cohort size, and the multiple variables that were assessed. Unlike prospective studies, retrospective studies do not have the benefit of standardizing data collection and therefore are subject to additional confounding factors and bias. However, as a pilot study at an Australian quaternary referral center, this study does utilize recent data, over a significant period of time (5 years); thus, attempting to provide a foundation for further studies. While larger cohort sizes are always preferred, this cohort of n = 250 has given statistically significant results. The multitude of variables included in this study may contribute to additional error in analysis. However, the primary intention of the authors was to determine which variables had the greatest associations with surgical margins and in doing so, direct future prospective studies towards more clinically relevant measures.

22 CONCLUSION This study has shown that the single most important variable in determining the size of the surgical margin in oral cavity cancer was the surgeon who performed the operation. High-volume surgeons delivered larger surgical margins. These findings support improved patient outcomes if surgery is performed at high-volume centers by high-volume surgeons. Future prospective research would provide addition evidence in the area. REFERENCES: 1. Mochizuki Y, Harada H, Ikuta M, et al.: Clinical characteristics of multiple primary carcinomas of the oral cavity. Oral Oncol 51:2, Weinstock YE, Alava I, 3rd and Dierks EJ: Pitfalls in determining head and neck surgical margins. Oral Maxillofac Surg Clin North Am 26:2, Luryi AL, Chen MM, Mehra S, et al.: Positive surgical margins in early stage oral cavity cancer: an analysis of 20,602 cases. Otolaryngol Head Neck Surg 151:6, Sutton DN, Brown JS, Rogers SN, et al.: The prognostic implications of the surgical margin in oral squamous cell carcinoma. Int J Oral Maxillofac Surg 32:1, Platz H, Fries R, Hudec M, et al.: The prognostic relevance of various factors at the time of the first admission of the patient. Retrospective DOSAK study on carcinoma of the oral cavity. J Maxillofac Surg 11:1, 1983

23 6. Eskander A, Merdad M, Irish JC, et al.: Volume-outcome associations in head and neck cancer treatment: a systematic review and meta-analysis. Head Neck 36:12, Vickers AJ, Bianco FJ, Serio AM, et al.: The surgical learning curve for prostate cancer control after radical prostatectomy. J Natl Cancer Inst 99:15, Martling A, Cedermark B, Johansson H, et al.: The surgeon as a prognostic factor after the introduction of total mesorectal excision in the treatment of rectal cancer. Br J Surg 89:8, Lee-Feldstein A, Anton-Culver H and Feldstein PJ: Treatment differences and other prognostic factors related to breast cancer survival. Delivery systems and medical outcomes. JAMA 271:15, Eastham JA, Kattan MW, Riedel E, et al.: Variations among individual surgeons in the rate of positive surgical margins in radical prostatectomy specimens. J Urol 170:6 Pt 1, Stojadinovic A, Leung DH, Hoos A, et al.: Analysis of the prognostic significance of microscopic margins in 2,084 localized primary adult soft tissue sarcomas. Ann Surg 235:3, Rahima B, Shingaki S, Nagata M, et al.: Prognostic significance of perineural invasion in oral and oropharyngeal carcinoma. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 97:4, Fagan JJ, Collins B, Barnes L, et al.: Perineural invasion in squamous cell carcinoma of the head and neck. Arch Otolaryngol Head Neck Surg 124:6, 1998

24 14. de Vicente JC, Rodriguez-Santamarta T, Rosado P, et al.: Survival after free flap reconstruction in patients with advanced oral squamous cell carcinoma. J Oral Maxillofac Surg 70:2, Hsieh TY, Chang KP, Lee SS, et al.: Free flap reconstruction in patients with advanced oral squamous cell carcinoma: analysis of patient survival and cancer recurrence. Microsurgery 32:8, Marchetti C, Pizzigallo A, Cipriani R, et al.: Does microvascular free flap reconstruction in oral squamous cell carcinoma improve patient survival? Otolaryngol Head Neck Surg 139:6, Chaturvedi P, Datta S, Nair S, et al.: Gross examination by the surgeon as an alternative to frozen section for assessment of adequacy of surgical margin in head and neck squamous cell carcinoma. Head Neck 36:4, Ord RA and Aisner S: Accuracy of frozen sections in assessing margins in oral cancer resection. J Oral Maxillofac Surg 55:7, Gerber S, Gengler C, Gratz KW, et al.: The impact of frozen sections on final surgical margins in squamous cell carcinoma of the oral cavity and lips: a retrospective analysis over an 11 years period. Head Neck Oncol 3: Chaturvedi P, Singh B, Nair S, et al.: Utility of frozen section in assessment of margins and neck node metastases in patients undergoing surgery for carcinoma of the tongue. J Cancer Res Ther 8 Suppl 1: Miyota S, Kobayashi T, Abe T, et al.: Intraoperative assessment of surgical margins of oral squamous cell carcinoma using frozen sections: a practical clinicopathological management for recurrences. Biomed Res Int 2014: 2014

25 Patient Characteristics N (%) or mean (SD) Cohort Size 250 (100%) DEMOGRAPHICS Age (yrs) 59.4 (SD: 13.90) SURGICAL Resective Surgeon (RS) RS Group 1 (>40 cases) 1 surgeon (7.7%); 122 cases (48.8%) RS Group 2 (20-40 cases) 3 surgeons (23.1%); 91 cases (36.4%) RS Group 3 (<20 cases) Reconstructive Surgeon Same Different Nil Reconstructive Type Nil Graft Locoregional Flap Free Flap Surgical Approach Per Oral Mandibulectomy Mandibulotomy Lip-split mandibulectomy Lip-split mandibulotomy Drop down Weber Ferguson 9 surgeons (69.2%); 37 cases (14.8%) 137 (54.8%) 51 (20.4%) 62 (24.3%) 62 (24.8%) 31 (12.4%) 45 (18.0%) 112 (44.8%) 187 (74.8%) 31 (12.4%) 3 (1.2%) 9 (3.6%) 8 (3.2%) 6 (2.4%) 6 (2.4%) Access Procedure 32 (12.8) HISTOLOGICAL Cancer Type Squamous cell carcinoma (SCC) Salivary carcinoma Sarcoma 237 (94.8%) 6 (2.4%) 7 (2.8%) Cancer Location Tongue 106 (42.4%) Buccal mucosa 16 (6.40%) Mandible 45 (18%) Floor of mouth 46 (18.40%) Retromolar trigone 14 (5.60%) Maxilla 23 (9.2%) Maximum tumor dimension 21.8 (SD: 12.33) T-stage T1 107 (42.8%) T2 68 (27.2%) T3 10 (4%) T4 65 (26%) Frozen Section performed 32 (12.8%) Lymphovascular invasion present 26 (10.4%) Perineural invasion present 60 (24.0%) Table 1: Summary of the predictor variables for the cohort SD = Standard deviation

26 Study Variable Coefficient (mm) 95% CI P-value DEMOGRAPHICS Age (yrs) to SURGICAL Resective Surgeon (RS) RS Group 1 (reference) RS Group 2 RS Group to to Reconstructive Surgeon Same (reference) Different Nil to to Reconstructive Type Nil (reference) Graft (omitted due to colinearity) Locoregional Flap Free Flap to to Surgical Approach Per Oral (reference) Mandibulectomy Mandibulotomy to to Lip-split mandibulectomy to Lip-split mandibulotomy Drop down Weber Ferguson to to to Access Procedure to HISTOLOGICAL Non-SCC to Cancer Location Tongue (reference) Buccal mucosa Mandible Floor of mouth to to to Retromolar trigone to Maxilla to Maximum tumor dimension to T-stage T1 (reference) T2 T3 T to to to Frozen Section performed to Lymphovascular invasion present to Perineural invasion present to Table 2: Regression analysis showing effect of predictor variables on the surgical margin

27 Table3: Mean surgical margin per resective surgeon volume group Resective Surgeon (RS) mean surgical margin (mm) 95% CI (mm) RS Group to RS Group to RS Group to Table 4: Beta-coefficient of variables with the surgical margin, in descending order of greatest impact Study Variable Beta-coefficient P - value RS Group2 vs RS Group RS Group3 vs RS Group Retromolar trigone Non-SCC Lip-split mandibulectomy Perineural invasion