Prognostic Implications of Skin Invasion in Locally Advanced Oral Cavity Squamous Cell Carcinoma

Abstract

Objective.

To evaluate the effect of histopathologic skin invasion on 2- and 5-year disease-free survival (DFS) and overall survival (OS) in patients treated with primary surgery for locally advanced oral cavity squamous cell carcinoma (OCSCC).

Study Design.

A retrospective case-control study was performed comparing previously untreated patients with pT4a OCSCC with and without skin invasion.

Setting.

Academic medical center.

Methods.

Propensity score–matched cohorts were derived by age, sex, surgical margins, pathologic N classification, adjuvant treatment, and primary tumor site. The Kaplan-Meier method was used to evaluate 2- and 5-year OS and DFS, which were compared between cohorts via the log rank (Mantel-Cox) test statistic.

Results.

Overall 25 patients were identified to have pathologic skin invasion, and 50 were selected for the matched control group. OS was significantly lower for patients with skin invasion as compared with controls at 2 years (30.8% vs 53.3%, P = .018) and 5 years (16.6% vs 42.2%, P = .01). DFS was significantly lower for patients with skin invasion vs controls at 2 years (23.7% vs 47.7, P = .037) and 5 years (15.8% vs 41.4%, P =.024).

Conclusion.

Histopathologic skin invasion in OCSCC is associated with dismal prognosis in patients who underwent primary surgical treatment. OS outcomes for patients with skin invasion are comparable to survival of patients with recurrent/metastatic disease and T4N2 disease.

Squamous cell carcinoma represents >90% of malignancies in the oral cavity worldwide. The diagnosis is often delayed with many patients presenting with advanced stage disease. Over the past few decades, many studies have identified that the presence of certain pathologic features, such as perineural invasion (PNI), lymphovascular invasion, multiple positive lymph nodes, extracapsular extension, and positive margins, is associated with higher risk of recurrence and death.^1–3 The presence of these high-risk features is used to recommend how patients are treated after primary surgery. Most patients who present with advanced-stage disease and high-risk features are recommended primary surgery, followed by adjuvant radiation and chemotherapy.⁴ Despite recent advances in combined modality therapies, approximately half of patients with high-risk features will develop recurrence.⁵

Skin invasion through direct extension of oral cavity squamous cell carcinoma (OCSCC) has been characterized as a form of dermal metastasis. Prior studies have suggested that skin invasion in locally advanced OCSCC is associated with worse prognosis than identical-stage disease without skin invasion.⁶ However, these studies investigated skin invasion in combination with other adverse features in OCSCC. Liao et al examined the prognostic impact of histopathologic risk factors in patients with T4a disease and reported that tumor invading mandibular cortex and/or skin had worse 5-year disease-specific survival (79% without invasion vs 69% with invasion) and 5-year overall survival (OS; 65% without invasion vs 52% with invasion). While significant, their outcome measures did not evaluate the independent effect of skin invasion on prognosis.⁷ Other studies investigated skin invasion as an independent risk factor for recurrence and death but were limited to patients with squamous cell carcinoma of the buccal mucosa. In these studies, skin invasion predicted significantly worse 5-year locoregional control and OS.^8,9 To date, no studies have independently investigated the effect of skin invasion on survival in patients with locally advanced OCSCC. Understanding the prognostic implications of skin invasion is crucial in considering adverse pathologic features of advanced OCSCC, predicting prognosis, and recommending adjuvant treatment plans.

The objective of this study was to evaluate the effect of histopathologic skin invasion on 2- and 5-year disease-free survival (DFS) and OS in patients treated with primary surgery for locally advanced OCSCC.

Methods

Data Source

Approval was obtained by the Institutional Review Board at the University of Kansas Medical Center. A retrospective case-control study was conducted including patients with locally advanced OCSCC undergoing primary surgical resection of disease from a single academic tertiary medical center between January 2004 and April 2019. Charts were identified by the research data repository of the Healthcare Enterprise Repository for Ontological Narration. The search was conducted by ICD-9 and ICD-10 (International Classification of Diseases, Ninth Revision and Tenth Revision) to identify patients diagnosed with malignancies of the oral cavity and by Current Procedural Terminology code to identify patients treated with excision of the disease with head and neck free tissue transfer flaps: 141.0-141.9 (ICD-9, tongue), 145.0-145.9 (ICD-9, other oral cavity), 144.0-144.9 (ICD-9, floor of mouth), 143.0-143.9 (ICD-9, gum), 140.0-140.9 (ICD-9, lip), C01-C02.9 (ICD-10, tongue), C05.0-C06.9 (ICD-10, other oral cavity), C04.0-C04.9 (ICD-10, floor of mouth), C03.0-C03.9 (ICD-10, gum), C00.0-C00.9 (ICD-10, lip); 15756 (free muscle or myocutaneous flap), 15757 (free skin flap), 15758 (free fascial flap), 15842 (free muscle flap for facial nerve paralysis), 20962 (free bone flap other than fibula, iliac crest, or metatarsal), 20969 (free osteocutaneous flap other than iliac crest, metatarsal or great toe), 20955 (free bone flap, fibula), 20955 (free bone flap, fibula), 20956 (free bone flap, iliac crest), 20970 (free osteocutaneous flap, iliac crest), and 43496 (free jejunum transfer).¹⁰ Data were extracted from the electronic health record and entered into a REDCap data collection form (Research Electronic Data Capture).¹¹

Patient Population

A retrospective case-control study was performed comparing previously untreated patients with pT4a OCSCC with and without skin invasion. Only patients with skin involvement by the primary tumor were included in the skin invasion population; nodal skin invasion was not included in this group. Patients aged ≥18 years with pT4a OCSCC treated with primary surgery were eligible for inclusion. Patients were excluded if they had distant metastatic disease or received chemotherapy, radiation therapy, or immunotherapy prior to surgical extirpation of disease. Additionally, patients who underwent surgery within 2 years prior to data analysis (surgery after April 2019) were excluded to ensure a minimum 2 years of clinical follow-up. Patients were also excluded if there were insufficient data in the electronic medical record.

Outcomes

Primary outcome measures were OS and DFS at 2 and 5 years. OS was defined as time to death from any cause from initial date of diagnosis. DFS was defined as time after primary surgery until date of recurrence or death, whichever occurred first. Both definitions were based on those from the National Cancer Institute (National Institutes of Health).^12,13 Other variables collected were patient demographics, primary tumor subsite, clinical tumor stage, adjuvant treatment type, and recurrence pattern, as well as key dates, such as diagnosis, surgery, start of adjuvant therapy, recurrence, and death. For each patient, pathologic slides were reviewed by 2 board-certified pathologists (S.W. and A.F.) to determine the presence or absence of skin invasion and other pathologic variables. Skin invasion was defined as direct extension by the primary tumor to the dermis or epidermis, and tumors were further categorized by tumor present within the lower half of the dermis (deep dermis), within the upper half of the dermis (superficial dermis), or in the dermis and epidermis. Other pathologic variables collected during the comprehensive review included pathologic T classification, tumor thickness, depth of invasion, tumor size, histologic grade, bone invasion, resection margin status, lymphovascular invasion, PNI, extranodal extension, skin invasion, type of skin invasion (deep dermis, superficial dermis, or dermis and epidermis), skin diameter, and number of positive lymph nodes.

Statistical Analysis

Data were analyzed with SPSS version 26 (IBM) and R version 3.5.1 (R Foundation for Statistical Computing). Propensity score–matched cohorts were created through the R package MatchIt.¹⁴ Propensity scores were derived by age, sex, surgical margins, pathologic N classification, and adjuvant treatment. A nearest-neighbor selection approach was utilized with 2:1 optimal ratio matching. The 50 closest matches were selected from a control cohort of 135 patients. Group comparisons of categorical data were performed with chi-square or Fisher exact tests (cell count <10). Group comparisons of scale variables were performed with Mann-Whitney U tests. Kaplan-Meier survival curves were developed for 2- and 5-year OS and compared via the log rank (Mantel-Cox) test statistic. Post hoc Cox regression analysis was performed to assess outcomes while adjusting for skin invasion, bone invasion, and PNI. In this analysis, all lip primary tumors were eliminated.

Results

A total of 172 patients met inclusion criteria. After pathologic re-review of all cases, 25 (14.5%) patients were identified as having pathologic skin invasion, and 50 patients without pathologic skin invasion were selected as a control group via propensity score matching. There were no significant differences in matched covariates between cases and controls. Patient demographics are described in Table 1.

Table 1.

Patient Demographics.^a

	Control (n = 50)	Skin invasion (n = 25)	Total (N = 75)	P value
Age,b y	63 (57-70)	62 (58.25-69.5)	63 (56.75-70)
Sex				.607
Men	34 (68)	17 (68)	51 (68)
Women	16 (32)	8 (32)	24 (32)
Margin				.158
Positive	6 (12)	6 (24)	12 (16)
Negative	44 (88)	19 (76)	63 (84)
Site of primary tumor				—
Mucosal lip	0 (0)	5 (20)	5 (6.7)
Buccal mucosa	4 (8)	2 (8)	6 (8)
Upper alveolar ridge	0 (0)	0 (0)	0 (0)
Lower alveolar ridge	15 (30)	7 (28)	22 (29.3)
Retromolar trigone	8 (16)	1 (4)	9 (12)
Floor of mouth	14 (28)	10 (40)	24 (32)
Tongue	9 (18)	0 (0)	9 (12)
Hard palate	0 (0)	0 (0)	0 (0)
Pathologic N stage				.398
N0	28 (56)	10 (40)	38 (50.7)
N1	3 (6)	1 (4)	4 (5.3)
N2	5 (10)	2 (8)	7 (9.3)
N3	14 (28)	12 (48)	26 (34.7)
Adjuvant treatment				.929
None	12 (24)	8 (32)	20 (26.7)
Chemotherapy	0 (0)	0 (0)	0 (0)
Radiation therapy	21 (42)	8 (32)	29 (38.7)
Chemotherapy and radiation therapy	13 (26)	6 (24)	19 (25.3)
Chemotherapy and immunotherapy	1 (2)	1 (4)	2 (2.7)
Radiation therapy and immunotherapy	2 (4)	1 (4)	3 (4)
Chemotherapy, radiation therapy, and immunotherapy	1 (2)	1 (4)	2 (2.7)
Perineural invasion				.030c
Present	21 (42)	17 (68)	38 (51)
Absent	29 (58)	8 (32)	37 (49)
Extranodal extension				.151
Yes	33 (67)	13 (52)	46 (62)
No	16 (33)	12 (48)	28 (38)

^aValues are presented as No. (%) unless noted otherwise.

^bMedian (interquartile range).

^cP < .05.

OS was significantly lower among patients with skin invasion vs those without at 2 years (30.8% vs 53.3%, P = .018) and 5 years (16.4% vs 42.2%, P = .01). DFS was also significantly lower for patients with skin invasion than for controls at 2 years (23.7% vs 47.7, P = .037) and 5 years (15.8% vs 41.4%, P = .024). Table 2 shows the complete survival data. Median survival time was approximately 1 year (11.5 months) greater for patients without skin invasion (21 months; 95% CI, 0-66) as compared with patients with skin invasion (9.5 months; 95% CI, 3.9-9.8). Figure 1 and Figure 2 show the Kaplan-Meier curves for OS and DFS. Table 3 shows a post hoc multivariable analysis to evaluate the effect of skin invasion on survival while controlling for bone invasion and PNI. After controlling for these potential covariates, skin invasion remained significantly associated with 2- and 5-year OS. The finding that PNI, but not skin invasion, was significantly associated with 2- and 5-year DFS is consistent with prior literature demonstrating the strong association between PNI and disease recurrence.

Table 2.

Survival Outcomes at 2 and 5 Years.

	Alive, % (95% CI)a
Survival	Control	Skin invasion	P value
Overall
2 y	53.3 (38.6-68)	30.8 (12.4-49)	.018
5 y	42.2 (26.9-57.5)	16.4 (0-34.4)	.010
Disease-free
2 y	47.7 (33.4-62)	23.7 (26.3-56.5)	.037
5 y	41.4 (26.3-56.5)	15.8 (0-33)	.024

^a95% CI = percentage alive ± 1.96 × SE.

Figure 1.

Kaplan-Meier overall survival (OS) curves: 2- and 5-year OS for patients with pT4a oral cavity squamous cell carcinoma with pathologic skin invasion and without (control).

Figure 2.

Kaplan-Meier disease-free survival (DFS) curves: 2- and 5-year DFS for patients with pT4a oral cavity squamous cell carcinoma with pathologic skin invasion and without (control).

Table 3.

Post Hoc Cox Regression Analysis.^a

	Skin invasion		Bone invasion		PNI
	P value	HR (95% CI)	P value	HR (95% CI)	P value	HR (95% CI)
OS
2 y	.037b	2.113 (1.046-4.270)	.064	2.792 (0.944-8.263)	.0001b	6.049 (2.791-13.106
5 y	.018b	2.224 (1.145-4.321)	.067	2.474 (0.938-6.523)	.0001b	5.049 (2.529-10.082)
DFS
2 y	.142	1.664 (0.843-3.283)	.110	2.107 (0.845-5.255)	.0001b	5.264 (2.562-10.812)
5 y	.069	1.838 (0.954-3.541)	.086	2.192 (0.894-5.371)	.0001b	4.397 (2.251-8.588)

Abbreviations: DFS, disease-free survival; OS, overall survival; PNI, perineural invasion.

^aPost hoc Cox regression analysis to evaluate effect of skin invasion on survival while controlling for bone invasion and PNI. Lip cases were excluded (n = 58). When bone invasion and PNI are controlled for, skin invasion remained statistically significant with 2- and 5-year OS. When skin invasion and bone invasion are controlled for, PNI remains statistically significant for 2- and 5-year OS and 2- and 5-year DFS.

^bP < .05.

The histopathologic distribution of skin invasion was as follows: 3 had invasion into deep dermis; 9 had invasion into superficial dermis; and 13 had invasion into dermis and epidermis. There was no significant difference in survival outcomes and type of histopathologic skin invasion. Figure 3 shows hematoxylin and eosin slides of OCSCC invading skin at the 3 depths. Within the skin invasion population, 20 of the 25 patients died; only 3 were confirmed alive; and 2 were lost to follow-up shortly after treatment of disease. Of 20 patient deaths, 14 were OCSCC or treatment related: 5 had locoregional recurrence; 5 developed distant metastatic disease; and 4 were discharged to hospice for progressive clinical deterioration without documented recurrence. Of the 20 patients who died, 4 were lost to follow-up after treatment, and cause of death was not documented. Of the 20 deaths, 2 were likely related to other factors, such as multiple traumatic hip fractures and progressive cognitive decline (n = 1) and pneumoperitoneum resulting in sepsis (n = 1).

Figure 3.

Hematoxylin and eosin slide of oral cavity squamous cell carcinoma invading skin at various depths: (a) deep dermal invasion, (b) superficial dermal invasion, and (c) dermal and epidermal invasion.

Discussion

This retrospective case-control study evaluated the prognostic effect of skin invasion in patients who underwent primary surgical treatment for pT4a OCSCC. Our results indicate that histopathologic skin invasion in surgically resected specimens was associated with a dismal prognosis, with 2- and 5-year OS (30.8% and 16.4%, respectively) and DFS (23.7% and 15.8%, respectively) comparable to or lower than survival outcomes seen in high-risk features and recurrent/metastatic OCSCC. The 5-year OS in our control population (42%) was consistent with OS reported in the literature, which ranges from 31.2% to 47%.^7,15–17 Median survival time was approximately 1 year (11.5 months) greater for patients without skin invasion (21 months) vs patients with skin invasion (9.5 months). These results highlight that the abysmal survival outcomes in patients with skin invasion should be used to guide conversations between patients and their care teams.

The aggressive nature of squamous cell carcinoma with skin involvement was first recognized by Cole and McGuirt in 1995. In this cohort of 26 patients, those with direct skin involvement had a median survival time of 7 months.⁶ Multiple studies since have concluded that skin involvement in head and neck squamous cell carcinoma is aggressive and portends poor prognosis. However, it is unclear why. A few studies hypothesized that skin invasion likely spreads along subdermal lymphatics, which leaves residual microscopic disease in the lymphatic system. In a retrospective study conducted by Fang et al, patients with buccal mucosal squamous cell carcinoma had lower 3-year locoregional control (41% vs 74%), and multivariate analysis showed that tumor invasion to skin of the cheek was the only prognostic factor that predicted worse locoregional control (P = .0014). They hypothesized that this subdermal lymphatic spread was likely the reason for low rate of locoregional control.⁸ In a retrospective analysis, Mair et al reviewed the cases of 125 patients with buccal mucosa squamous cell carcinoma who received primary surgery and bilateral neck dissections and investigated patterns affecting contralateral neck disease. Multivariate analysis showed that ipsilateral nodal metastasis and skin involvement were both independently predictive of contralateral nodal metastasis. This group hypothesized that skin involvement was due to spread through the submucosal plexus, which would ultimately lead to lymphatic spread to ipsilateral and contralateral cervical micrometastases.¹⁸ Although the mechanism of direct skin extension is incompletely understood, our data and the literature show that skin invasion is a prognostic feature that portends a poor prognosis.

The 5-year rates for OS (16.4%) and DFS (15.8%) in our patients with OCSCC skin invasion are comparable to previously published survival outcomes for patients with T4 OCSCC with other high-risk features, such as extracapsular extension, PNI, lymph node metastasis, and recurrent/metastatic OCSCC. In a retrospective study of 137 patients with T4 OCSCC treated by surgery and radiotherapy, Tankere et al found that 5-year OS rates were 37.7% in patients with N1 disease and 15.8% in patients with N2 disease.¹⁹ In a retrospective study with 181 patients with pT4 OCSCC treated with radical resection, 5-year DFS rates were 35% for patients with extracapsular extension and 35% for patients with PNI.⁷ Of note, survival time in OCSCC with skin invasion is comparable to patients with recurrent and metastatic OCSCC. Although our patients with direct skin invasion had a higher median survival (9 months) than those with recurrent/metastatic head and neck squamous cell carcinoma treated only with chemotherapy (6.6 months) as seen in KEYNOTE-040,²⁰ median survival was lower than those treated with pembrolizumab alone (14.9 months) or pembrolizumab with chemotherapy (13 months) in KEYNOTE-048.²¹ Our results highlight the abysmal survival rates of pT4a OCSCC presenting with skin invasion, which should guide initial counseling of patients and families.

Current standard of care for patients with pT4a OCSCC presenting with skin invasion includes curative surgical resection with adjuvant chemoradiation, which is unlikely to significantly extend survival. Patients with extensive skin involvement often suffer from severe pain, bleeding, odor, and chronic infection, which all contribute significantly to poor quality of life. Therefore, it is crucial that shared decision making among the patient, family, and a multidisciplinary team is emphasized. Goals for care targeting physical and psychosocial symptoms should be communicated clearly, and palliative care should be engaged early.²² A retrospective study at Mount Sinai investigated palliative surgery with pedicled and free flap reconstruction as an option for patients with head and neck cancer with extensive skin involvement. This study concluded that palliative surgery was able to completely remove the tumor in 96% patients, allowing patients to live their final months without suffering the morbidity of large fungating tumors. Similar to our study, median OS in this study was 9.5 months.²³ For certain patients with significant medical comorbidities and a desire for less aggressive treatment, primary palliative therapy should be the mainstay of treatment. Nilsen and Johnson discussed the burden on quality of life and inability to extend life associated with palliative surgery and/or chemoradiation treatments on patients with very advanced head neck cancer. Instead, they highlighted the importance of comprehensive palliative care and supportive services that are congruent with patients’ end-of-life goals.²⁴ Early engagement in goals-of-care discussions and palliative options is crucial to present to this population.

Our study is not without limitations. As this is a retrospective case-control study, there were patients who were lost to follow-up, for whom survival outcome data were incomplete. However, these patients were censored in the Kaplan-Meier analyses, which assumes that loss to follow-up is unrelated to mortality. Additionally, retrospective case-control studies are more susceptible to selection bias, as the control and case cohorts are representative of similar populations. Yet, selection bias can be addressed only in a randomized prospective study, which we believe to be impractical due to the low incidence of this disease presentation. In addition, our sample size is small, as our patient data were obtained from a single tertiary academic center and an initial presentation of disease with direct skin invasion is rare. PNI was statistically significant between our control group and skin invasion population. Due to the limited population, we considered it feasible to match cohorts based on age, sex, surgical margins, pathologic N classification, and adjuvant treatment. Post hoc the difference in PNI was significant. While this limits a decoupling of the effects of skin invasion vs PNI in the present analysis, it did find skin invasion to be a significant poor prognostic indicator, and future consideration with a larger cohort will need to be done to determine the effects of PNI vs skin invasion independently.

Conclusion

Patients with pT4a OCSCC who present with histopathologic skin invasion have significantly worse survival outcomes as compared with propensity score–matched controls without skin invasion. In fact, this patient population had 5-year OS and DFS rates comparable to patients with T4 OCSCC with high-risk features such as positive nodal disease, PNI, extracapsular extension, and even metastatic/recurrent OCSCC. As such, curative intent surgery with adjuvant chemoradiation should be reconsidered the standard of care in this population. Palliative care and supportive services that are congruent with patients’ end-of-life goals should be engaged early in diagnosis. Alternative palliative surgical approaches would be viable options to decrease significant morbidity associated with skin involvement.