Staging system performance and clinical outcomes for cutaneous squamous cell carcinoma of the ear: A single center retrospective study

Abstract

Background:

Cutaneous squamous cell carcinoma (cSCC) of the ear is associated with poor outcomes. No studies have evaluated current staging system performance in this specific location.

Objective:

Describe clinicopathologic characteristics and outcomes of ear cSCC and evaluate the performance of current staging systems.

Methods:

Retrospective study including cases diagnosed and treated at a cancer center from January 2000 to December 2014. Demographic, clinical, and pathological data was collected from clinical records. Biopsy slides were re-reviewed and patients were staged according to AJCC 7th, 8^th and BWH staging.

Results:

Of 125 patients, the mean age at diagnosis was 71.9 years (SD 12.5), with a majority of males (89.6%, n=112). Median follow-up was 22.3 months. Local recurrence and survival risk factors were similar to cSCC outside the ear. The Akaike’s Information Criterion (AIC) estimates showed that the BWH system better predicted outcomes than the AJCC 7th, and the AJCC 8th, with AIC values of 189.9, 270.5 and 274.1, respectively.

Limitations:

Retrospective design, single center study, no control group.

Conclusion:

Current staging systems perform well at stratifying risk in ear cSCC.

Introduction:

The majority of cutaneous squamous cell carcinoma (cSCC) carries an excellent prognosis, with a low risk of distant or nodal metastasis (1.1% - 5.2%), and low disease-specific death (1.5% - 2.1%), based on recent population-based studies.^1–8 However, a subset of cSCC, termed “high-risk” cSCC (HRcSCC) has an increased risk of developing metastases to both regional lymph nodes and distant sites.^{6, 9, 10} Both patient factors and features of the primary tumor contribute to this elevated risk.^{6, 9–11} These known risk factors include patient immunosuppression, large tumor size, deep invasion, poor differentiation, perineural or lymphovascular invasion, desmoplasia, history of prior treatment, and certain anatomical locations such as the non-hair bearing lip and the external ear.^9–12

Ear cSCC represents 11-25% of cSCCs in males and 0.2 – 3% in females.¹³ cSCC of the ear has been associated with worse prognosis in multiple studies.^{12, 14, 15} The American Joint Committee on Cancer (AJCC) 7^th edition cSCC staging included ‘location on the ear or lip’ as a high-risk feature for tumor (T) classification that upstaged tumors.¹⁶ While tumor location on the ear and lip have been found to be independent and significant risk factors for metastasis and disease-specific death,^{2, 10} the quality of evidence has been limited by the size of the reported datasets along with the non-systematic nature of the case selection. The 2017 AJCC 8^th Edition featured important changes due to demonstrated shortcomings of the 7^th Edition in accurately stratifying disease outcomes.^{17, 18} One of these changes was the removal of special anatomic sites (the ear and lip location), from the T classification system. Furthermore, the specific features of ear tumors that denote higher risk of poor outcomes have not been well-described.

A recent retrospective cohort study found that the AJCC 8^th Edition T classification system was superior to the 7^th edition in both homogeneity (depicting similar outcomes within categories) and monotonicity (portraying worse outcomes with increasing categories).¹⁹ Also, the Brigham Women’s Hospital (BWH) staging system, developed based on tumor features, has shown good homogeneity and monotonicity for cSCC.²⁰ While several retrospective studies have demonstrated relatively poor outcomes of patients with advanced cSCC of the ear, there is little published information on the characteristics and prognosis of ear SCC across all stages of disease as well as the performance of current staging systems on this anatomic location.^21–23 The objective of this study was to describe clinicopathologic characteristics and outcomes of cSCC located on the ear and to evaluate the performance of current staging systems on cSCC that arise in this unique anatomic site.

Patients and methods:

This was a retrospective study performed at Memorial Sloan Kettering Cancer Center and approved by the Institutional Review Board. All biopsy-proven invasive cSCCs of the ear treated from January 1^st, 2000 to December 31^st, 2014 were identified. Exclusion criteria included cases that were not treated at our institution (i.e., the patient presented for a consult visit and ultimately sought treatment elsewhere), in situ SCC, tumors that did not originate from the ear (i.e. primary tumors of the parotid), primary tumors of the internal ear or ear canal, and less than 6 months follow up (except for death before 6 months, which was included).

Demographic data:

Data was extracted from clinical, operative, and pathology reports by 3 investigators (C.N-D., S.M., K.C.). Demographic and clinical patient data collected included age, sex, history of radiation or traumatic scar at the tumor site, and immunocompromised status (defined as: solid organ transplant, hematologic malignancy, HIV, applicable chemotherapy, immunosuppressive medication). Primary tumor characteristics (largest tumor diameter, primary/recurrent tumor) as well as radiologic or pathologic presence of lymph node or visceral metastases at diagnosis were documented.

Histopathological analysis:

All diagnostic biopsy and excision specimens’ slides were re-reviewed for consensus by four authors (C.N.-D., S.M., K.C., and K.J.B.) including a board-certified dermatopathologist (K.J.B) for complete histopathologic characterization of the tumors. This review was done to update for tumor information that was not recorded at the time of initial review (e.g. depth, etc). Histopathologic features collected included differentiation (well, moderate, or poor), keratoacanthomatous or desmoplastic features, perineural invasion (PNI, including size of nerve caliber [≥0.1 mm]), lymphovascular invasion (LVI; including the extent of invasion), tumor depth (in mm) and Clark’s level (I to V).

Staging, treatment, and follow-up:

Each tumor was staged according to AJCC 7^th and 8^th Edition, as well as BWH if the complete record of contributory variables was available. The type of primary treatment provided to the tumor (wide local excision, Mohs micrographic surgery, definitive radiation therapy, topical treatments, or observation), nodal basin lymphadenectomy, adjuvant radiation therapy, and any available imaging modality information were recorded. Development of local recurrence, lymph node or distant metastasis, as well as overall survival (OS) and disease-specific survival (DSS) were recorded. The duration of follow-up was also recorded.

Statistical analysis:

Descriptive and relative frequencies, means, and standard deviations were used to describe the study participants, lesion characteristics, surgical procedures, staging, and outcomes information. Chi-square statistics were used to evaluate the distribution of study variables by locoregional recurrence (LRR) of SCC. For survival analysis, follow-up time was calculated from the date of surgical procedure to the last known follow-up date or the date of death. Estimates were created for overall and disease-specific follow-up. Kaplan-Meier estimates and Cox proportional hazards models were created. Since the three staging systems use overlapping components, three separate Cox models were estimated. Each model included patient sex (female = 0, male = 1) and locoregional recurrence (no = 0, yes = 1). Proportionality assumptions were visually assessed via log-log plots and by testing of scaled Schoenfeld residuals. To compare the performance of the individual staging systems, Akaike’s Information Criterion (AIC) was calculated after fitting each of the Cox models. Lower AIC scores denote a model with better performance. All analyses were performed using Stata/MP 16.1, Stata Corporation, College Station, Tx.

Results

From 2000 to 2014, 227 patients presented to our institution with cSCC of the ear; of these, 147 were evaluated and completely treated at our institution. Twenty-two cases were excluded (12 corresponded to other tumor types, 7 had missing data, and 3 were collision tumors). Final analysis included 125 invasive cSCC.

Patient characteristics:

Mean age at diagnosis was 71.9 years (SD 12.5). Most patients were males (89.6%, n=112). A total of 16 patients (12.8%) were immunosuppressed; 11 (8.8%) being solid organ transplant patients (Table 1).

Table 1.

Patient, lesion, and treatment characteristics by whether the patient had recurrent disease (n=125).

Variable	Coding	Overall (n=125)		Locoregional recurrence		p-value
				No (n=105)	Yes (n=20)
		n (%)		n (%)	n (%)
Sex	Female	13 (10.4)		8 (7.6)	5 (25)	0.02
	Male	112 (89.6)		97 (92.4)	15 (75)
Differentiation	Poor	27 (21.6)		18 (17.1)	9 (45)	0.05
	Moderate	46 (36.8)		39 (37.1)	7 (35)
	Well	34 (27.2)		32 (30.5)	2 (10)
	Not recorded	18 (14.4)		16 (15.2)	2 (10)
Pre-treatment size (largest diameter)	<2.0 CM	65 (52.0)		60 (57.1)	5 (25.0)	0.02
	≥ 2.0 CM	22 (17.6)		15 (14.3)	7 (35.0)
	Not recorded	38 (30.4)		30 (28.6)	8 (40.0)
Keratoacanthoma features	No	50 (78.1)		38 (36.2)	12 (60.0)	0.04
	Yes	14 (21.9)		14 (13.3)	0 (0)
	Not recorded	61 (48.8)		53 (50.5)	8 (40.0)
Perineural invasion (PNI)	No	51 (40.8)		43 (72.9)	8 (44.4)	0.03
	Yes	26 (20.8)		16 (27.1)	10 (55.6)
	Not recorded	48 (38.4)		46 (43.8)	2 (10.0)
PNI nerve caliber	<0.1	2 (7.7)		1 (6.2)	1 (10.0)	0.76
	>0.1	13 (50.0)		8 (50.0)	5 (50.0)
	Not recorded	11 (42.3)		7 (43.8)	4 (40.0)
LVI	No	67 (53.6)		56 (53.3)	11 (55.0)	0.016
	Yes	13 (10.4)		7 (6.7)	6 (30.0)
	Not recorded	45 (36.0)		42 (40.0)	3 (15.0)
Clark level	1	2 (1.6)		2 (1.9)	0 (0)	0.03
	2	17 (13.6)		16 (15.2)	1 (5.0)
	3	10 (8.0)		10 (9.5)	0 (0)
	4	4 (3.2)		4 (3.8)	0 (0)
	5	53 (61.6)		36 (34.3)	17 (85.0)
	Not recorded	39 (31.2)		37 (35.2)	2 (10.0)
Margins involved	No	15 (93.8)		11 (91.7)	4 (100)	0.55
	Yes	1 (6.3)		1 (8.3)	0 (0)
	Total	16 (100)		12 (100)	4 (100)
Immune status	Not involved	104 (83.2)		87 (82.9)	17 (85)	0.75
	Not recorded	5 (4)		3 (2.9)	2 (10)
	Solid organ transplant	11 (8.8)		10 (9.5)	1 (5)
	Lymphoma	1 (0.8)		1 (1)	0 (0)
	HIV	1 (0.8)		1 (1)	0 (0)
	Chemotherapy	1 (0.8)		1 (1)	0 (0)
	Immunosuppressed	2 (1.6)		2 (1.9)	0 (0)
Special features	No	108 (86.4)		94 (89.5)	14 (70)	0.006
	Prior RT	13 (10.4)		7 (6.7)	6 (30)
	Over scars	4 (3.2)		4 (3.8)	0 (0)
Recurrent disease	No	99 (79.2)		88 (83.8)	11 (55)	0.004
	Yes	26 (20.8)		17 (16.2)	9 (45)
Lymph node on imaging	No	51 (40.8)		39 (37.1)	12 (60)	0.004
	Yes	20 (16)		14 (13.3)	6 (30)
	Not imaged	54 (43.2)		52 (49.5)	2 (10)
Treatment	None	1 (0.8)		1 (1)	0 (0)	<0.001
	Mohs surgery	57 (45.6)		56 (53.3)	1 (5)
	WLE	59 (47.2)		44 (41.9)	15 (75)
	Topical treatment	1 (0.8)		1 (1)	0 (0)
	Radiation therapy	1 (0.8)		0 (0)	1 (5)
	Chemotherapy	6 (4.8)		3 (2.9)	3 (15)
Adjuvant radiation	No	116 (92.8)		101 (96.2)	15 (75)	0.001
	Yes	9 (7.2)		4 (3.8)	5 (25)
LN excision with surgery	No	94 (75.8)		84 (80.0)	10 (50)	0.003
	Yes	30 (24.2)		20 (19.0)	10 (50)
	Not recorded	1 (0.80)		1 (1.0)	0 (0.0)
LN involved by tumor	No	19 (15.2))		13 (12.4)	6 (30)	0.79
	Yes	11 (8.8)		7 (6.7)	4 (20)
	Not recorded	95 (76.0)		85 (81.0)	10 (50.0)
Extranodal extension	No	27 (84.4)		20 (87)	7 (77.8)	0.52
	Yes	5 (15.6)		3 (13)	2 (22.2)
	Not recorded	93 (74.4)		82 (78.1)	11 (55.0)
AJCC V.7	1	67 (53.6)		62 (59.1)	5 (25.0)	0.001
	2	40 (32.0)		31 (29.5)	9 (45.0)
	3	6 (4.8)		5 (4.8)	1 (5.0)
	4	7 (5.6)		3 (2.9)	4 (20.0)
	Not recorded	5 (4.0)		4 (3.8)	1 (5.0)
Brigham’s woman hospital (BWH) staging	T1	33 (51.6)		31 (60.8)	2 (15.4)	0.004
	T2a	9 (14.1)		8 (15.7)	1 (7.7)
	T2b	10 (15.6)		5 (9.8)	5 (38.5)
	T3	12 (18.8)		7 (13.7)	5 (38.5)
	Unable to stage	61 (48.8)		54 (51.4)	7 (35.0)
	Total	125 (100)		105 (100)	20 (100)
AJCC V.8	1	77 (61.6)		71 (67.6)	6 (30.0)	0.001
	2	13 (10.4)		11 (10.5)	2 (10.0)
	3	24 (19.2)		17 (16.2)	7 (35.0)
	4	7 (5.6)		3 (2.9)	4 (20.0)
	Not recorded	4 (3.2)		3 (2.9)	1 (5.0)
LN metastasis	No	55 (44.0)		48 (45.7)	7 (35.0)	<0.001
	Yes	12 (9.6)		4 (3.8)	8 (40.0)
	Not recorded	58 (46.4)		53 (50.5)	5 (25.0)
Distant metastasis	No	56 (44.8)		47 (44.8)	9 (45.0)	0.005
	Yes	11 (8.8)		5 (4.8)	6 (30.0)
	Not recorded	58 (46.4)		53 (50.5)	5 (25.0)
Status	Alive	79 (63.2)		71 (67.6)	8 (40.0)	0.02
	Dead	46 (36.8)		34 (32.4)	12 (60.0)
	Unknown	0 (0)		0 (0)	0 (0)
Cause of death	SCC-related	18 (39.1)		8 (23.5)	10 (83.3)	0.001
	Not SCC-related	28 (60.9)		26 (76.5)	2 (16.7)

Tumor characteristics:

Of the 125 ear cSCC, 79.2% (n=99) were primary tumors and 20.8% (n=26) were recurrent. The median diameter was 12 mm (95% CI 10 – 15; min/max: 2-140 mm). Most tumors were moderately differentiated (36.8%, n=46). Median tumor depth (mm) was 5 mm (IQR:12.7, 0.6 – 50mm). Approximately one-third (n=26, 33.8%) presented with PNI and 16.3% (n=13) with LVI. Radiologic lymph node involvement was seen in 16% (n=20) of cases.

Survival analysis:

Median follow-up for the cohort was 22.3 months (IQR: 44.2; 1 – 177 months). At the time of chart review, 63.2% (n=79) of patients were alive. There were 46 deaths (36.8%), of which 18 (38.8%) were due to cSCC. Factors associated with OS and DSS are presented in Table 1. Increasing age (p=0.017 and <0.001), presence of LVI (p<0.001 and 0.001), prior RT (0.06 and 0.008), recurrent tumor (p=0.001 and 0.001), lymph node on imaging (p=0.02 and 0.012), lymph node excision (p=0.001 and <0.001), T3/4 on AJCC7th (p<0.001 and <0.001), BWH T2b/T3, T3/4 AJCC 8^th (p<0.001 and 0.001), lymph node metastasis (p<0.001 and <0.001), and LRR (p<0.001 and <0.001) were all factors associated with poor OS and DSS (Table S1–Supplemental Digital Content).

Staging systems performance:

To evaluate and compare the performance of the existing staging systems (BWH, AJCC 7^th ed., AJCC 8^th ed.) for overall survival of cSCC of the ear, separate Cox proportional hazards models were developed. The results of these analyses are presented in Table 2 and Figure 1. The three systems presented comparable estimates for LRR and patient sex (male). The AJCC systems had similar hazard ratios for T2 and T3/4 disease. For the BWH system, T3 disease has the highest observed hazard ratio (HR), with these patients being 6.1 times more likely to die of disease over the course of follow-up compared to T1 patients. To compare model performance, Akaike’s Information Criterion (AIC) estimates indicated the BWH system performed better than the AJCC 7^th, and the AJCC 8^th, with AIC values of 189.9, 270.5 and 274.1, respectively

Table 2:

Overall and disease-specific survival based on different staging systems.

			Overall survival		Disease specific survival
Staging system	Stage	N (%)	HR (95%)	p-value	HR (95%)	p-value
AJCC V.7	T1	68 (56.2)	1.0 (referent)	--	1.0 (referent)	--
	T2	40 (33.1)	1.53 (0.70-3.37)	0.28	1.64 (0.37-7.36)	0.517
	T3/4	13 (10.7)	6.54 (2.50-17.13)	<0.001	15.85 (4.22-59.52)	<0.001
BWH	T1	33 (51.6)	1.0 (referent)	--	1.0 (referent)	--
	T2a	9 (14.1)	1.53 (0.48-4.84)	0.471	Does not estimate
	T2b	10 (15.6)	3.73 (1.33-10.51)	0.013	11.20 (1.15-108.62)	0.037
	T3	12 (18.8)	8.47 (2.94-24.34)	<0.001	45.44 (5.24-393.9)	<0.001
AJCC V.8	T1	77 (63.6)	1.0 (referent)	--	1.0 (referent)	--
	T2	13 (10.7)	1.08 (0.24-4.70)	0.92	Does not estimate	--
	T3	24 (19.8)	3.74 (1.71-8.17)	0.001	9.19 (2.18-38.77)	0.003
	T4	7 (5.8)	13.53 (4.23-43.31)	<0.001	45.30 (9.84-208.49)	<0.001

^*AJCC: American Join Committee on Cancer; BWH: Brigham and Women’s Hospital.

Figure 1.

Coefficient plot for the hazard ratio for death in patients with cSCC of the ear. Three separate models were estimated, one for each of the cSCC staging systems. Each model includes estimates for locoregional recurrence (yes), sex (male) and staging system levels.

Treatment modality and recurrence:

The most common initial treatment of cSCC on the ear was wide local excision (WLE) in 47.2% (n=59) followed by Mohs micrographic surgery (MMS) in 45.6% (n=57) of cases. Adjuvant radiation was used in 7.2% (n=9) of cases. A sentinel lymph node biopsy was performed in 24.2% (n=30) of cases and was positive for cSCC in 36.7% (n=11) of those examined; extranodal extension was found on 5 of 11 (45.5%). After definitive therapy, 12 patients (17.9%) developed lymph node metastasis and 11 patients (16.4%) developed distant metastasis. Patients with a local recurrence had a worse OS and DSS (Figure 2).

Figure 2.

Kaplan-Meier estimates of overall (A) and disease specific survival (B) by locoregional recurrence status.

When comparing different treatment modalities, patients with ear cSCC treated with MMS had lower rates of LRR when compared with other treatment modalities. Only 1 out of 57 (1.7%) patients recurred with MMS vs 15 out of 59 with WLE (25.4%; p<0.001). The HR of recurrence was 3.8 (1.9 – 7.9) and 6.23 (2.14 – 18.1) for WLE and other treatments when compared to MMS (p<0.001 and 0.001, respectively). However, tumor characteristics differed significantly in patients treated with WLE compared to Mohs surgery. Compared to WLE, patients that received MMS had a higher proportion of well-differentiated tumors (33% vs 10.2%) and fewer poorly differentiated tumors (7% vs 33.9%, respectively). Tumors that were treated with MMS were smaller than those treated with WLE (mean 5.7mm vs 17.3mm). Finally, cases treated with MMS had a lower T-stage compared to cases treated with WLE (Table S2–Supplemental Digital Content).

Discussion

Ear cSCC has historically been considered a high-risk anatomic location, with location on the ear previously included as a factor for upstaging in the AJCC staging system. One hypothesis for worse prognosis of cSCC arising on the external ear is the relative lack of subcutaneous tissue in this anatomic location. Thus, cSCC arising on the ear is in close contact with dermal or subcutaneous lymphatics and cartilage early in the disease course which may lead to a higher risk of developing regional disease and metastasis.¹³

In this retrospective review of 125 invasive cSCC of the ear, we found several predictors of LRR, OS, and DSS, such as tumor depth, grade of differentiation, and presence of PNI or LVI. Importantly, the factors associated with negative outcomes in cSCC of the ear were similar to well-established risk factors for cSCC that arise in other anatomic locations. This study confirms that these accepted risk factors can improve risk-stratification of patients presenting with ear cSCC as they do in cSCC that arises in other anatomic locations. A notable strength of our study compared to other published retrospective studies is that all available histopathological slides were re-reviewed to assess for specific and updated pathologic features of each tumor. This systematic evaluation allowed for the accurate assessment of tumor risk factors that led to poor outcomes in our cohort of ear cSCC.

A recent study by Ruiz et al. comparing AJCC 7th, AJCC 8th, and BWH staging systems on 459 cSCC patients found that BWH staging would minimize inappropriate upstaging of low-risk disease. However, ear cSCC was not included in their cohort.²⁰ In our study, we found that BWH staging discriminated well between low-risk ear tumors (T1/T2a) and higher-risk tumors (T2b/T3). The AJCC 8th edition staging also provided good discriminatory data between different staging groups. In contrast, we found that the AJCC 7th edition led to increased upstaging of low-risk (T2) tumors. When comparing staging systems using AIC, BWH performed better than AJCC 8th and 7^th edition for ear cSCC, similarly to the Ruiz et al. findings on non-ear cSCC. Performance of different staging systems for cSCC of the external ear has not been previously assessed.

Treatment modalities for cSCC of the ear include Mohs micrographic surgery, WLE, and radiation therapy. In our study, about half of the patients were treated with standard WLE and half with Mohs surgery. Patients with ear cSCC treated with Mohs surgery had a lower loco-regional recurrence when compared with other treatment modalities. This might be in part explained due to better local control with Mohs vs WLE or other treatments.²⁴ However, our results, and any study showing Mohs superiority, should be interpreted critically. In our cohort, Mohs cases tended to have a better degree of differentiation, smaller clinical size, and lower T-stage. Additionally, since our study and most published studies are retrospective, there might be additional factors that were used to decide a particular treatment and there is a high risk for selection bias. A recent study by Gonzalez et al. evaluated Mohs surgery on ear cSCC. They found a recurrence rate of 4.1% (7 out of 140 cases) after a mean time of 11.9 months.¹³ Six out of these 7 patients (85.7%) with recurrent disease died of cSCC after a mean time of 22.7 months, confirming that local recurrence is an important risk factor affecting survival. We saw similar results in our study, where recurrent tumors had a worse prognosis. In the same study, 18 out of 170 patients (10.6%) developed nodal metastasis during follow-up; 44% died of disease after presenting with nodal metastasis.¹³ van Lee et al compared recurrence rates of cSCC treated with Mohs vs WLE and found almost 3 times reduction in recurrence rate with Mohs (3% vs 8%).²⁵ Finally, Xiong et al. found that Mohs surgery compared with WLE in T2a cSCCs was associated with lower recurrence rates and fewer adverse outcomes (defined as local recurrence and disease-specific death).²⁶

Limitations:

The main limitations of our study are its relatively small sample size. However, we only included cases that had the pathology slides available for re-review for complete pathological variables, given the evolving field of SCC staging over years. Also, in situ SCC cases were excluded, reducing the total number of patients included. Additionally, the study was retrospective, single center, and the lacked a control cohort including cSCC of other anatomic sites. The next step would be to compare ear cSCC to other sites. Moreover, these patients were seen in a highly specialized tertiary cancer center and were treated by both dermatologists and head and neck surgeons as part of a multidisciplinary skin cancer management tumor board.²⁷ It is likely that these tumors were more advanced than those typically seen in general dermatology clinics. For example, 33.8% of the cohort presented with PNI and 16.3% with LVI. In the case of the analysis for the BWH staging system, some cases lacked information on the clinical variables needed for complete staging. Nevertheless, the models were similar with regards to covariates except for the staging system.

Conclusion:

This study confirms that cSCC of the ear can behave aggressively with locoregional metastasis as well as distant disease and death. We found that the variables that predict poor outcomes in non-ear cSCC are the same for ear cSCC. Therefore, current staging systems perform well in stratifying risk for patients presenting with cSCC on this special site. Future population-wide studies are needed to determine whether ear cSCCs are associated with poor outcomes compared to cSCC in other locations.