A study was undertaken to determine whether bone invasion is an independent prognostic factor in oral squamous cell carcinoma (SCC) after taking into account the extent of bone invasion.
A study was undertaken to determine whether bone invasion is an independent prognostic factor in oral squamous cell carcinoma (SCC) after taking into account the extent of bone invasion.
The study was a retrospective review of 498 patients with oral SCC undergoing surgery with curative intent, 102 of whom had pathologically proven bone invasion. Bone invasion was categorized as absent, cortical, or medullary and tested for association with disease control and survival.
After adjusting for potential confounding factors in multivariate analysis, there was no association between cortical invasion and overall (P = .48) or disease-specific survival (P = .63). In contrast, medullary invasion was an independent predictor of reduced overall (hazard ratio [HR], 1.9; 95% confidence interval [CI], 1.2-3.1; P = .006) and disease-specific survival (HR, 2.1; 95% CI, 1.2-3.6; P = .01), and this appeared to result from an increased risk of distant metastatic failure (P = .037) rather than local (P = .51) or regional recurrence (P = .14). Within the subset of patients with medullary invasion, survival differed significantly according to tumor size (P = .029).
Patients with oral SCC and bone invasion have widely variable outcomes depending on the depth of bone invasion and tumor size. The results suggest that the current American Joint Committee on Cancer staging system, which classifies all tumors invading through cortical bone as T4, has limited prognostic utility. The authors recommend a revision of the T staging system such that tumors are classified as T1 to T3 based on size and then upstaged by 1 T stage in the presence of medullary bone invasion. Cancer 2011. © 2011 American Cancer Society.
Squamous cell carcinoma (SCC) of the oral cavity frequently invades the mandible. Since the first edition of the Manual for Staging of Cancer was published in 1977,1 bone invasion has remained as 1 of the criteria by which T4 tumors are defined, albeit with the more recent caveat that superficial erosion is inadequate to classify gingival primaries as T4.2 This designation was based on the consensus opinion of the committee members that bone invasion heralds a poor clinical outcome.
Subsequent studies examining the prognostic significance of bone invasion have shown conflicting results. Whereas some authors have reported an association with adverse outcomes on univariate analysis,3-7 other studies did not support this finding.8-10 Furthermore, multivariate analyses taking into account potential confounding factors such as tumor size and involved margins have suggested that bone invasion is not an independent predictor of survival.4-6 Unfortunately, the majority of the studies to date were limited by small sample sizes and failed to distinguish between cortical and medullary bone invasion, making reliable conclusions difficult.
Our experience suggests that patients with small tumors and clear resection margins can expect favorable outcomes, even in the presence of bone invasion.11 We hypothesized that the poor prognosis attributed to bone invasion reflects an association with other adverse prognostic features such as tumor size and involved surgical margins. The primary aim of this study was to determine whether bone invasion is an independent prognostic factor in oral SCC after taking into account the extent of bony invasion and potential confounding variables. The secondary aim was to analyze the prognostic utility of the current American Joint Committee on Cancer (AJCC) T staging system for patients with oral SCC and bone invasion.
The Sydney Head and Neck Cancer Institute at Royal Prince Alfred Hospital, Sydney, Australia, maintains a prospective clinicopathological database. After ethics approval from the Area Health Service review board, a retrospective search was carried out to identify patients with oral SCC undergoing primary surgical treatment with curative intent between 1987 and 2009.
Bone invasion was categorized as absent, cortical when limited to the cortex, or medullary when extension into cancellous bone was present. The pathology reports of all tumors exhibiting bone invasion were reviewed. In cases where there was inadequate information to determine the extent of bone invasion (n = 44), the stored pathology slides were reviewed by a pathologist experienced in the examination of head and neck tumors (R.M.). In patients where the slides were not available for review and the extent of bony invasion was unknown, these cases were initially analyzed separately and then combined with the cortical group for the final multivariate models on the grounds of parsimony, because there was no effect on other covariates regardless of how they were categorized.
Statistical analysis was performed using Stata, version 9.0 (Stata Corporation, College Station, Tex). Cross-tabulated data were analyzed using chi-square test or Fisher exact test where appropriate. All statistics were 2-sided, and a value of P < .05 was considered statistically significant. Overall survival was calculated from the date of surgery to date of death or last follow-up. For disease-specific survival, patients who died from causes other than oral SCC were censored at the time of death. Univariate Cox proportional hazards regression was used to test the association of bone invasion with overall survival, disease-specific survival, local recurrence, regional recurrence, and distant metastatic recurrence. Other potential covariates, including age at diagnosis, sex, tumor size (analyzed as both a continuous and a stratified variable: tumor ≤2 cm, >2 cm and ≤4 cm, >4 cm), tumor differentiation, involved surgical margins (clear, close [<5 mm], involved), perineural invasion, lymphovascular invasion, extracapsular nodal spread (absent, microscopic, macroscopic), and N stage (N0, N1, N2a, N2b, N2c, N3), were also tested for prognostic significance. Statistically significant and clinically important covariates were used to create a multivariate Cox proportional hazards regression model to determine whether bone invasion is an independent prognostic factor. Transformation of continuous variables was performed where appropriate, and proportional hazards assumptions were tested. Survival curves were generated using PASW Statistics 17.0 (SPSS Inc., Chicago, Ill) by both the Kaplan-Meier method and proportional hazard models to demonstrate the estimated effect of bone invasion on recurrence and survival.
The study population consisted of 498 patients with primary oral SCC treated by surgical resection, with 377 undergoing concurrent neck dissection. There were 311 men and 187 women, with a median age of 63.5 years (range, 22.5-91.5 years) and median follow-up of 39.3 months. Adjuvant radiotherapy was administered in 205 patients with a median dose of 56 grays (Gy) to the primary site (range, 12.8-70 Gy) and 50 Gy to the neck (range, 12.8-70 Gy). Relevant patient demographic and clinicopathological data are summarized in Table 1. Pathologically proven bone invasion was present in 102 patients. Within this group, 41 (40.2%) had invasion limited to the cortex, whereas 61 (59.8%) had medullary involvement.
|Floor of mouth||152||30.5|
|Close (<5 mm)||242||48.6|
|Pathological T stage|
|Pathological N stage|
|Extracapsular nodal spread|
|Yes, <60 Gy||135||27.1|
|Yes, ≥60 Gy||70||14.1|
There was a significant association between bone invasion and tumor size when analyzed as a continuous variable, with mean maximal diameters of 2.4 cm, 3.3 cm, and 3.8 cm in those with absent, cortical, and medullary bone invasion, respectively (P < .001). Similar results were obtained when tumor size was analyzed as a stratified categorical variable (P < .001). There was also a significant association between bone invasion and perineural invasion (P = .012), lymphovascular invasion (P = .001), tumor differentiation (P = .045), involved surgical margins (P < .001), N stage (P = .047), and extracapsular nodal spread (P = .019). These associations were stronger for medullary than cortical invasion.
The median overall survival in the study population was 10.4 years, with 149 deaths, of which 95 were because of oral SCC. As demonstrated in Figures 1A and B, patients with medullary bone invasion had a significantly reduced 5-year overall survival of 29%, compared with 65% in those without bone invasion (P < .001). Similarly, 5-year disease-specific survival rates were 36% in the presence of medullary invasion and 77% with no bone invasion (P < .001). Patients with cortical invasion alone had similar overall (P = .463) and disease-specific survival (P = .824) compared with those without bone invasion (Table 2).
|Variable||Overall Survival||Disease-Specific Survival|
|Univariate Analysis||Multivariate Analysisa||Univariate Analysis||Multivariate Analysisa|
|HR (95% CI)||P||HR (95% CI)||P||HR (95% CI)||P||HR (95% CI)||P|
|Age, ≥65 vs <65 y||2.3 (1.7-3.2)||<.001||2.1 (1.5-3.0)||<.001||2.3 (1.5-3.5)||<.001||2.0 (1.3-3.1)||.002|
|Sex, female vs male||0.9 (0.8-1.1)||.215||0.9 (0.8-1.1)||.472|
|Close [<5 mm] vs clear||1.5 (1.0-2.2)||.040||1.1 (0.7-1.7)||.667||2.0 (1.2-3.3)||.007||1.3 (0.7-2.2)||.392|
|Involved vs clear||3.4 (2.2-5.2)||<.001||2.1 (1.3-3.4)||.002||4.2 (2.4-7.4)||<.001||2.0 (1.1-3.7)||.030|
|Moderate vs well||1.3 (0.8-2.1)||.261||1.8 (0.9-3.3)||.078|
|Poor vs well||1.3 (0.7-2.4)||.372||1.9 (0.8-4.2)||.123|
|Perineural invasion||2.0 (1.4-2.9)||<.001||2.5 (1.7-3.9)||<.001|
|Lymphovascular invasion||2.2 (1.4-3.5)||.001||3.1 (1.8-5.1)||<.001|
|Tumor size, stratified||<.001||<.001|
|>2 cm and ≤4 cm vs ≤2 cm||1.7 (1.2-2.4)||.007||1.2 (0.8-1.8)||.335||2.2 (1.3-3.6)||.002||1.6 (0.9-2.7)||.102|
|>4 cm vs ≤2 cm||3.6 (2.2-5.7)||<.001||1.8 (1.1-3.2)||.030||6.3 (3.6-11.3)||<.001||2.9 (1.5-5.7)||.002|
|Cortical vs absent||0.8 (0.4-1.5)||.463||0.8 (0.3-1.7)||.478||0.9 (0.4-2.0)||.824||0.8 (0.3-2.0)||.626|
|Medullary vs absent||3.0 (2.0-4.7)||<.001||1.9 (1.2-3.1)||.006||3.3 (2.0-5.5)||<.001||2.1 (1.2-3.6)||.010|
|N stage||<.001||1.2 (1.0-1.4)||.010||<.001||1.3 (1.1-1.6)||.001|
|N1 vs N0||1.5 (0.9-2.4)||.143||1.8 (0.9-3.5)||.079|
|N2a vs N0||0.6 (0.1-4.0)||.551||1.3 (0.2-9.6)||.791|
|N2b vs N0||2.4 (1.6-3.6)||<.001||3.5 (2.1-5.7)||<.001|
|N2c vs N0||4.4 (2.5-7.6)||<.001||7.2 (3.9-13.2)||<.001|
|N3 vs N0||19.0 (5.8-62.0)||<.001||34.9 (10.4-117.8)||<.001|
|Extracapsular nodal spread||<.001||<.001|
|Microscopic vs absent||2.6 (1.7-4.0)||<.001||1.2 (0.7-2.0)||.536||3.7 (2.3-5.9)||<.001||1.2 (0.6-2.1)||.617|
|Macroscopic vs absent||4.4 (2.1-9.0)||<.001||2.4 (1.0-5.8)||.046||6.1 (2.8-13.5)||<.001||3.0 (1.1-8.0)||.031|
In addition to medullary bone invasion, significant predictors of reduced overall and disease-specific survival on univariate analysis were age ≥65 years, involved or close surgical margins, perineural and lymphovascular invasion, N stage, extracapsular nodal spread, and tumor size (Table 2). There was no difference in overall (P = .339) or disease-specific survival (P = .905) when comparing patients undergoing marginal versus segmental mandibulectomies.
Multivariate survival analysis was performed using purposeful selection of covariates as per Hosmer et al,12 with the final models shown in Table 2. After adjusting for the effect of tumor size and other covariates including adjuvant radiotherapy, patients with cortical invasion did not have any significant difference in overall (P = .478) or disease-specific survival (P = .626) when compared with patients without bone invasion. In contrast, medullary invasion was a significant independent predictor of reduced overall (hazard ratio [HR], 1.9; 95% confidence interval [CI], 1.2-3.1; P = .006) and disease-specific survival (HR, 2.1; 95% CI, 1.2-3.6; P = .010) (Fig 1C and D). Similar results were obtained if age and tumor size were considered as continuous variables in the analysis. When bone invasion was simply classified as present or absent, it was not a significant predictor of overall (P = .184) or disease-specific survival (P = .203). There were no significant interactions between tumor size and type of bone invasion.
There were 69 documented local recurrences and 71 regional recurrences occurring with a median time to recurrence of 11 and 15 months, respectively. Although there was weak evidence for an association between medullary bone invasion and local (P = .060) and regional failure (P = .046) on univariate analysis, this association was not maintained on multivariate analysis (P = .507 and P = .143, respectively).
Distant metastases were the sites of first recurrence in 23 patients. This occurred in 9.8% of patients with medullary bone invasion, 4.9% of those with cortical invasion, and 3.8% of those with no bone invasion. Univariate analysis revealed a 330% increased risk of distant metastases when comparing patients with medullary bone invasion versus those without bone invasion (HR, 4.3; 95% CI, 1.7-11.3; P = .003). After including clinically and statistically significant covariates in a multivariate analysis, this relationship remained significant (HR, 3.0; 95% CI, 1.1-8.1; P = .037). Other independent predictors of distant failure included advanced N stage (P < .001) and increasing tumor size (P = .008).
As shown in Figure 2A, the current AJCC T staging system does not stratify disease-specific survival outcomes well in our study population, with significant overlap between T3 and T4 tumors. As size and medullary bone invasion are independent prognostic factors, an alternative T staging system was examined in which tumors are classified as T1 to T3 based on size and then upstaged by 1 T stage in the presence of medullary bone invasion (Table 3). The criteria for T4 designation unrelated to bone invasion (such as extrinsic tongue muscle involvement or unresectable disease) would remain unchanged. This proposal reflected our finding that medullary invasion increases the risk of death regardless of tumor size, but as size had a greater effect, patients with small tumors and medullary invasion should not be automatically upstaged to T4. Furthermore, patients with medullary bone invasion demonstrated widely variable survival outcomes based on tumor size (Fig. 3). Survival outcomes according to the revised T staging system are shown in Figure 2B.
|Primary Tumor (T)||Definition|
|Tx||Primary tumor cannot be assessed|
|T0||No evidence of primary tumor|
|Tis||Carcinoma in situ|
|T1||Tumor ≤2 cm in greatest dimension and no medullary bone invasion|
|T2||Tumor >2 cm but not >4 cm in greatest dimension and no medullary bone invasion or tumor ≤2 cm in greatest dimension with medullary bone invasion|
|T3||Tumor >4 cm in greatest dimension and no medullary bone invasion or tumor >2 cm but not >4 cm in greatest dimension with medullary bone invasion|
|T4a||Tumor >4 cm in greatest dimension with medullary bone invasion or tumor invades into deep (extrinsic) muscle of tongue (genioglossus, hyoglossus, palatoglossus, and styloglossus), maxillary sinus, or skin of face|
|T4b||Tumor involves masticator space, pterygoid plates, or skull base and/or encases internal carotid artery|
The AJCC staging system for oral SCC continues to classify tumors with invasion through cortical bone as T4, which by definition assigns the patient to stage IV disease with resultant prognostic and management implications.2 This is despite current evidence suggesting that bone invasion is not an independent prognostic factor when confounding variables such as tumor size and involved surgical margins are taken into consideration.4-6, 9 However, the studies to date are generally limited by small sample sizes and failure to distinguish between cortical and medullary invasion, making it difficult to derive any definitive conclusions. The present study reconciles the difference between prior studies and expert opinion by demonstrating that only medullary bone invasion is an adverse prognostic feature after adjusting for potential confounding variables in a study with sufficient power.
Central to our hypothesis was the assumption that the association between bone invasion and survival related more to advanced tumor size, rather than any particular biological effect of bone invasion. Our analysis supported this in part, by confirming that large tumors (>4 cm) have approximately triple the risk of cancer-related death (HR, 2.9; 95% CI, 1.5-5.7; P = .002) and that cortical bone invasion had no association with recurrence or survival. Despite this, we were forced to reject our initial hypothesis because of the significantly increased risk of cancer-specific death identified in patients with medullary bone invasion (HR, 2.1; 95% CI, 1.2-3.6; P = .010), after adjusting for the effect of size and other relevant covariates. These findings differ from those of previous studies that distinguished between cortical and medullary invasion, but found no prognostic significance on multivariate analyses.5, 10 This may reflect differences in covariate selection and limited statistical power (type II error), as both studies had small sample size in comparison with the current analysis.
Interestingly, when bone invasion was classified as simply present or absent, it was not a significant independent predictor of overall (P = .225) or disease-specific survival (P = .221). This is in keeping with previous studies that classified bone invasion in this way and found no association with adverse outcomes.4, 6, 8, 9 The exception to this was a retrospective analysis by Jones et al of 82 patients with oral and oropharyngeal SCC undergoing mandibulectomy.7 However, 31 (93.9%) of 33 patients with bone invasion had oral cancer in this study, compared with 17 (34.7%) of 49 patients without bone invasion, introducing primary site as a potential confounding factor.
Despite affirming the importance of medullary bone invasion, our results challenge the current AJCC staging system, which classifies all patients with bone invasion through the cortex into a single high-risk category. It is problematic to recommend that all tumors with medullary invasion be classified as T4, because the TNM staging system implies a stratification of risk from T1 through T4. Our results suggest that tumor size is a more important predictor of survival than medullary invasion. Therefore, the predicted hazard would be higher, for example, in a 5 cm tumor without bone invasion compared with a 2 cm tumor with extension to medullary bone. Furthermore, medullary invasion increased the risk of death regardless of tumor size. This is supported by the finding that there was no interaction between tumor size and the type of bone invasion to suggest that the effect of medullary invasion varies according to the size of the primary. Finally, patients with medullary invasion had widely variable survival outcomes based on tumor size (Fig. 3). The proposed alternative method of T staging outlined in Table 3, in which tumors are classified as T1 to T3 based on size and then upstaged by a T stage in the presence of medullary bone invasion, provides better stratification of patient prognosis (Fig. 2B). Although this model performed well in our study population, it clearly needs to be validated by large independent patient cohorts.
Although our findings support the current literature suggesting that cortical bone invasion alone does not adversely affect prognosis, we are cautious to make recommendations regarding the omission of adjuvant radiotherapy. We found no difference in locoregional control rates based on the presence or absence of cortical invasion; however, adjuvant radiotherapy was administered to 73% of patients with cortical invasion compared with only 34% of patients without bone invasion. These differences make direct comparison difficult, and the similar disease control rates may be because of the effect of radiotherapy. Although a subgroup analysis of patients not receiving adjuvant radiotherapy suggested survival was not compromised in the presence of cortical invasion, this needs to be confirmed by other institutions.
Having established that medullary bone invasion is an independent predictor of reduced survival, we sought to investigate the patterns of recurrence in this group of patients. Contrary to our expectations, there was no association between medullary invasion and locoregional failure, which is in keeping with previously reported findings.5, 6, 8, 9 However, a significant association between medullary bone involvement and risk of distant metastases as the site of first recurrence and treatment failure clearly exists. This may be because of hematogenous dissemination of tumors that gain access to the circulation via the cancellous bone of the mandible, or alternatively may reflect underlying aggressive tumor biology manifesting as both medullary invasion and distant spread.
In this study, 10% of patients with medullary invasion presented with distant metastases as the site of first recurrence. This likely underestimates the true incidence, because our database does not record distant metastases in the presence of locoregional failure, and also patients may be referred directly to a palliative care physician in this setting. Assuming that future studies validate our results and confirm that a significant proportion of treatment failures in patients with medullary bone invasion are because of distant metastases, these patients should be considered for adjuvant systemic therapy in the context of a prospective trial.
In this retrospective single-institution analysis, we demonstrated that tumor size and medullary bone invasion are independent predictors of reduced survival. In contrast, tumors with bone invasion limited to the cortex have a similar prognosis to those without bone invasion. The adverse effect of medullary involvement appears to result, at least in part, from an increased risk of distant metastatic failure. Our results suggest the current AJCC T staging system, which classifies all such tumors as T4, has limited prognostic utility. If our results are validated by studies on independent patient cohorts, we recommend a revision of the T staging system.
No specific funding was disclosed.
The authors made no disclosures.