Prognostic Factors in Oral Squamous Cell Carcinoma Treated with Primary Bone Resection

Harper L. Wilson, M.D.1, William L. Beuerlein, M.D.2, Javier A. Laurini, M.D.2, Neelima Tummala, M.D.3, Gregory B. Russell, M.S.4, and Joshua D. Waltonen, M.D.1

Article Table of Contents  
Abstract Introduction Methods Statistical Analysis Results Discussion


This study aims to determine prognostic factors for survival and recurrence in cases of oral squamous cell carcinoma (SCC) undergoing surgery including bone resection by examining patient characteristics, pathological findings, and treatment variables including the type of bone resection. This analysis is a retrospective chart review of 84 adult patients with oral SCC treated with rim or segmental mandibulectomy or maxillectomy. Patient characteristics, pathologic findings, and treatments were collected by chart review and analyzed for correlation with outcome measures of overall survival, disease-specific survival, and local, regional, and distant recurrences. Review of pathologic slides was performed for the presence and extent of bone invasion. Median follow-up was 28 months with a median overall survival of 10 months. Medullary bone invasion was detected in 41 patients (49%) and cortical bone invasion in 2 patients (2%). Predictors of overall survival (OS) in univariate analysis were tumor grade (p=0.03), tumor size (p=0.003), tumor thickness (p=0.002), lymphovascular invasion (LVI) (p=0.0006), perineural invasion (PNI) (p=0.04), N stage (p<0.0001), and type of bone resection (p=0.01). T stage (p=0.058) and medullary bone invasion (p=0.070) neared significance. Only N stage remained significant on multivariate analysis (p=0.009). N stage predicted regional (p=0.003) and distant (p=0.006) recurrence. Tumor grade (p=0.007), tumor thickness (p=0.012), and LVI (p=0.0003) also predicted for distant recurrence. Nodal stage is a stronger clinical prognosticator for survival and recurrence in oral SCC than patient or tumor factors including tumor stage, bone invasion, and type of bone resection. Patients with known nodal metastases may benefit from more extensive neck dissection, intensified adjuvant therapy, and more rigorous surveillance imaging schedules.


Bone invasion is a frequently observed event in the pathogenesis of oral squamous cell carcinoma (SCC). Subsites with increased proximity to the mandible such as the alveolar ridge, retromolar trigone (RMT), and floor of mouth (FOM) have an increased risk of bone invasion1. Bone invasion through the cortical bone of the mandible or maxilla contributes to staging and is classified as T4a in the American Joint Committee on Cancer (AJCC) v7 staging system.2-3 Staging is a significant factor in determining management of oral cavity cancer, which typically includes primary surgery with or without bone resection and adjuvant radiation therapy (RT) or concurrent chemoradiotherapy (CRT) as indicated by pathology results.4

Conflicting literature leaves the prognostic value of bone invasion in oral SCC in question. Some studies argue that bone invasion is not a prognostic factor4-6 while some suggest otherwise.7-10 Reasons for these disparate findings may include failure to control for the depth or pattern of invasion, differences in type of surgery, and presence of other factors such as cervical lymph node metastases. Determining prognostic factors in cases of oral SCC with potential bone invasion is important to providing patients with a more informed prognosis and guiding adjuvant therapy recommendations.

Our primary aim is to examine the potential prognostic factors in oral SCC treated with primary surgery including bone resection and determine whether patient characteristics, pathologic findings, or treatment variables have a significant impact on survival and recurrence.


This study is a retrospective chart review of adult patients diagnosed with oral cavity SCC and treated with surgery that included a form of bone resection between January 2011 and September 2017 at the Wake Forest Baptist Medical Center. This study was approved by the hospital’s Institutional Review Board. Diagnoses were confirmed through histological examination of tumor tissue slides. Patient characteristics, operative variables, pathologic variables, and outcome measures were collected by chart review. Exclusion criteria included histology other than SCC, carcinoma in-situ, and incomplete documentation. Outcome measures collected from chart review included overall survival (OS), disease-specific survival (DSS), time to recurrence, and type of recurrence (local, regional, or distant). Local disease recurrence was defined as within the ipsilateral side of the oral cavity, regional disease recurrence was defined as within the head and neck, and distant disease recurrence was defined as below the head and neck.

Tumors that extended into multiple contiguous subsites due to size were classified as the central site of the tumor. Staging and determination of margins was performed in accordance with the AJCC v7 TNM staging guidelines.3 Extent of bone invasion was identified through review of original histology slides of bone tissue. Two pathologists determined the extent of bone invasion with independent examination and were blinded to clinical outcomes. Patients lost to follow-up at this institution were queried in the Wake Forest Baptist Health Cancer Registry for last known follow-up, recurrences, treatments, and deaths.

Statistical Analysis

Descriptive statistics, including means and standard deviations for continuous measures and frequencies and proportions for categorical data, were calculated for all study measures. Subjects who were alive at their last known follow-up were censored as of that date for survival and censored for recurrence if recurrence-free at that point; OS was calculated as the interval between the date of surgery and date of death or last date known to be alive. DSS was censored at the time of death of those patients known to have died of non-disease causes. Time to recurrence (local, regional, distance) was calculated from surgery date to known recurrence date or last date known to be disease-free. The Kaplan-Meier method was used to estimate survival (overall survival and time-to-recurrence) data. Cox proportional hazards models were used to assess the association between survival endpoints and independent measures; hazard ratios and 95% confidence intervals were created for predictors. Univariate models were analyzed initially; variables with p-values <0.10 were included in an initial multivariate model. A best-fit model was created by dropping non-significant variables from the model in a singular fashion until only significant effects remained. P-values less than 0.05 were considered to be statistically significant. SAS (version 9.4, Cary, NC, USA) was used for all analyses.


Eighty-four patients met criteria for inclusion. Patient and tumor characteristics are included in Table 1. Fifteen patients had a prior history of head and neck cancer (18%) including 10 patients with a remote history of radiation for a previous primary tumor (12%). Adjuvant therapy was not recommended in 38 patients (45%), recommended and received in 28 patients (33%), and recommended but not received in 15 patients (17.9%).

squamous cell carcinoma table 1 patient tumour characteristics

Pathological Findings
Tumor size ranged from 0.4-10 cm with a median of 4 cm. Tumor thickness ranged from 0.2-4.8 cm with a median of 1 cm.

Clinical Outcomes
Postoperative follow-up ranged from 1-95 months with a median of 28 months. In patients still living at the time of this analysis, the median follow-up was 35 months. Indications for recommending adjuvant radiotherapy included bone invasion, lymphovascular invasion (LVI), perineural invasion (PNI), N2 or N3 stage; addition of chemotherapy was recommended for positive margins or extranodal extension (ENE). Sixty- three patients met criteria for adjuvant therapy; of these, 10 received RT and 16 received CRT. OS ranged from 0-74 months with a median of 10 months. Within 2 months of surgery, 5 patients died due to complications of disease and 1 due to unrelated causes. Time to recurrence ranged from 1-71 months with a median of 10 months. Including patients with multiple types of recurrences, 21 patients (25%) developed local recurrence, 16 (19%) developed regional recurrence, 12 (14%) developed distant recurrence, and 52 patients (62%) did not recur. 5 patients (6%) developed a second primary Head and Neck SCC.

Risk Factors for OS
Table 2 demonstrates analysis of risk factors for survival.

squamous cell carcinoma table 2 disease specific survival

In univariate analysis, significant predictors of OS were histological grade (p=0.033), tumor size (p=0.0031), tumor thickness (p=0.0024), LVI (p=0.0006), PNI (p=0.043), N stage (p<0.0001), and surgery type (p=0.010). On multivariate analysis in which histological grade, tumor size, LVI, PNI, ENE, medullary versus no bone invasion, T stage, N stage, and type of bone resection met criteria for inclusion, N stage (p=0.0094) remained a significant predictor. LVI (p=0.048), N stage (p=0.0011), and medullary versus no bone invasion (p=0.017) were significant predictors in the best-fit analysis. Age, sex, margins, ENE, T stage (p=0.058), smoking history, tumor subsite, and postoperative RT/CRT were not significant predictors of OS or DSS. See Figures 1-3.

Risk Factors for DSS
Histologic grade (p=0.016), tumor size (p=0.0078), tumor thickness (p=0.016), LVI (p<0.0001), PNI (p=0.0084), N stage (p=0.0001), and surgery type (p=0.029) were significant predictors of DSS in univariate analysis.

squamous cell carcinoma figure 1 2 t stage n stage 

squamous cell carcinoma figure 3 surgery type 

Medullary versus no bone invasion neared significance (p=0.051) in univariate analysis. N stage (p=0.0039) remained significant in multivariate analysis in which histological grade, tumor size, LVI, PNI, ENE, T stage, N stage, and type of bone resection were included. Predictors in the best-fit analysis were LVI (p=0.043), N stage (p=0.0008), and medullary versus no bone invasion (0.020).

squamous cell carcinoma figure 4 risk factors

Risk Factors for Local, Regional, and Distant Recurrence
Risk factors for recurrence were analyzed and are shown in Figure 4. No risk factors were significant for local recurrence, but N stage (p=0.057) trended toward significance. Regional recurrence was predicted by N stage (p=0.0025). Significant predictors identified for distant recurrence included histological grade (p=0.0073), tumor thickness (p=0.012), LVI (p=0.0003), N stage (0.0064), and postoperative CRT (p=0.035). Margins, tumor size, ECS/ENE, T stage, tumor subsite, smoking history, alcohol use, and surgery type were not significant predictors of recurrence.


This study is a single-institution retrospective review with the aim of evaluating prognosticators in oral SCC treated with bone resection. Controversy remains concerning the impact of primary tumor factors, staging, bone invasion, and type of bone resection. Further analysis is needed to inform prognosis and treatment and address these conflicting results.

Our analysis finds that the strongest predictor of outcomes following these surgeries is nodal stage. This was true for overall and disease-specific survival. Nodal stage is well established in literature as a reliable prognosticator of survival in oral SCC1.3,4,8,9,11,12 We found that soft tissue factors such as tumor size, tumor thickness, LVI, and PNI were prognosticators in univariate analysis, but nodal status was the only remaining significant predictor in multivariate survival analysis. While N+ status fared worse than N0 patients, we did not see significantly poorer outcomes in advanced nodal status (N2/N3) compared to N1 patients. Also notable was the lack of significance of ENE on prognosis, although this analysis could be influenced by the small number of cases with ENE (n=15).

We also found that nodal stage neared significance for prediction of local recurrence and became a significant predictor of regional and distant recurrence. The predictive value of nodal status for recurrence has been demonstrated in literature.11,13 Our finding of its influence specifically on distant recurrence is similar to the results of an analysis of 498 oral SCC patients by Ebrahimi et al3. In our study, the two strongest predictors of distant recurrence were LVI and nodal stage, supporting the established risk of lymphatic involvement for distant recurrence.

Interestingly, margin status was not a significant predictor in any of our analyses, which is consistent with a recent study of 96 patients with oral SCC by Fives et al1 but distinct from most literature.3,8-12 Potential reasons for this discrepancy in the significance of margins include difference in surgeon preference or institutional practice for the procurement of margins or lack of an adequate number of patients with positive margins (n=11 in this series, 8 of which had other risk factors as well). We also noted several cases where positive margins were seen in the “main” pathology specimens but additional margins obtained from the surgical defect were negative; we classified these as positive margins in our analysis and would seem to have a lower risk of recurrence than “true” positive margins than cannot be cleared. With the preponderance of evidence in the Head and Neck Oncology literature demonstrating high risk of recurrence with positive margins, we do not advise de-intensifying adjuvant therapy in patients with positive margins based on this portion of our analysis.

Analysis of T stage neared significance in our analysis but did not achieve it, which is similar to the findings of Fives et al1 but inconsistent with more literature to date.4,8,9 Over half of the patients in this study had advanced pathologic T staging (T3 or T4), with 11.9% and 42.9% of patients staged as T3 and T4, respectively (Table 1). In this study, we found 8 patients with T4 stage treated with surgery alone (they had no other risk factors and each declined recommended adjuvant therapy). Only one patient recurred locally, at 11 months. It is worth studying whether an adequate aggressive resection may be sufficient therapy for this particular group of patients with no other high-risk features.

Type of mandibulectomy was significant in univariate but not multivariate analysis of survival, which is supported by findings of Mucke et al in a study of 982 patients with oral SCC4. Lack of prognostic value is also demonstrated in other literature6,8,13 and argues against the importance of the extent and method of bone resection in determining prognosis. The preoperative judgement of the surgeons in these series using physical exam and imaging findings to decide on a rim vs. a segmental mandibulectomy appears to be adequate to determine the likelihood of complete tumor resection. Factors that influence this decision include signs of mandible erosion, size and location of tumor, and available remaining mandibular height below the resection.6

Though bone invasion has a significant impact on staging, controversy remains over the prognostic impact of the presence and depth of bone invasion. We aimed to examine this issue, but of 43 patients with bone invasion, invasion was medullary in 41 cases and cortical in only 2 (Supplemental Figures 1 & 2). This finding was confirmed by each pathologist independently and upon additional review.

squamous cell carcinoma supplemental figure 1 2 bone invasion 

It differs with a number of studies which found a higher instance of cortical bone invasion3,8,12,13 but is similar to the findings in a recent study by Fives et al.1 The underrepresentation of cortical bone invasion may be due to the use of drills or burs for marginal mandibulectomy in several instances, resulting in a lack of bone specimen in cases of potential cortical invasion. This may have also resulted in a few false positive margins, where the tumor and periosteum were “peeled” off the mandible, with a positive deep margin that was subsequently cleared by the surgeon’s drilling of the adjacent mandibular margin.

Our analysis is therefore unable to examine the influence of cortical bone invasion but may serve to examine the effect of medullary bone invasion. In that context, medullary bone invasion neared significance in univariate and multivariate analyses of survival but achieved significance in the best-fit analysis. Prognostic significance of medullary invasion is supported in findings of several other studies.1,3,6,11,12

In comparison with similar studies performed by Fried et al12 and Fives et al,1 we found that tumor size > 4cm was a poor prognosticator compared to tumors < 4cm without invasion, but there was no significant difference in survival between cases of <4 cm with or without invasion. This indicates a principal role of tumor size rather than invasion in survival, which is supported by the results of Fried et al12 and our finding that tumor size was a univariate predictor of survival.

We examined the influence of tumor differentiation through histologic grade and found significance in univariate but not multivariate analysis of survival, and there was a significant impact on distant but not locoregional recurrence. Okura et al14 showed similar findings in a study of 345 patients with lower gingival SCC, but more conclusive studies argue against the significance of grade.3,4,6,10 Further studies with increased sample sizes may be able to further investigate this finding.

We were surprised to find that while 63 patients had pathology findings meriting recommendations for adjuvant therapy, only 26 actually underwent RT or CRT. We found a number of reasons why the other 37 patients did not receive therapy. Seven had undergone prior irradiation to the oral cavity or oropharynx and were not deemed to be candidates (3 remain alive with no recurrence, 2 developed recurrences, 1 died of a carotid blowout a month later, and 1 was lost to follow-up). Eight other patients with advanced stage disease and multiple risk factors died in the 3-month postoperative period from medical and/or surgical complications including dehydration, malnutrition, failure to thrive, and sepsis from infection of non-healing wounds; 3 of these had already experienced recurrences by the time of their deaths. Similarly, 5 patients were regarded as too frail to withstand adjuvant therapy; all 5 recurred by 24 months. Three patients had a positive margin as the sole criterion but did not receive adjuvant therapy; 1 of these had a local recurrence at 20 months successfully salvaged, and all 3 are alive. Finally, 4 patients with multiple risk factors declined the recommended adjuvant therapy and all 4 recurred by 6 months.

Patients at high risk for postoperative medical or surgical complications (based on nutrition status, comorbidities, history prior RT) may be faced with delays in initiating adjuvant therapy. Other patients may present for adjuvant therapy with poor performance status, which may be present prior to surgery or may develop in the postoperative period. In those cases, when multiple high-risk features are present, prognosis appears to be quite grim. Likewise, when patients with multiple high-risk features decline therapy, the outlook seems to be dismal. Efforts should be made to reduce the likelihood of complications including proper optimization of patients in the preoperative setting15 and postoperative care pathways.16 Patients with advanced stage tumors, especially with nodal disease, should be counseled as to the dismal outcomes following surgery alone, and should clearly understand preoperatively the notion of multidisciplinary care and the need for adjuvant therapy.

Our analysis faces several limitations. This study is at risk for the inherent bias present in retrospective reviews and its single-institution manner leads to potential sampling bias such as the underrepresentation of cases of cortical bone invasion. The small number of patients reduces the power to determine statistical significance. Furthermore, the inclusion of patients with recurrent tumors, a history of head and neck cancer or radiation, and patients undergoing maxillectomy limits the intended applicability to cases with primary tumors potentially invading the mandible. Strengths of this study include the blinded nature of the pathologic slide review for depth and extent of invasion and the rigorous statistical analysis for both survival and recurrences.


This single-institution retrospective study demonstrates that the strongest prognostic factor in oral SCC requiring bone resection is nodal stage rather than patient factors, pathologic findings such as tumor stage, or the extent or type of surgical resection. For patients with multiple high-risk features who do not undergo indicated adjuvant therapy, prognosis is grim.

No financial support given. The authors report no conflicts of interest.