Epidemiology of Malignant Cutaneous Granular Cell Tumors: A United States Population-Based Cohort Analysis using the Surveillance, Epidemiology, and End Results (SEER) Database

Abstract

Background

Malignant cutaneous granular cell tumors (mcGCT) are rare and associated with substantial morbidity and mortality. The literature includes single-institution studies.

Objective

To examine the incidence, secondary malignancies, treatment, overall survival (OS), and disease-specific survival (DSS) of patients with mcGCT.

Methods

A population-based cohort analysis was conducted in the Surveillance, Epidemiology, and End Results (SEER) database, 1973–2013, for patients diagnosed with mcGCT. Risk-adjusted associations between OS/DSS and patient characteristics and treatment modalities were assessed using Cox proportional-hazard regression. Quantile regression determined median survival times.

Results

Five-year DSS was 62.8%. Patients demonstrated an increased risk for renal and pancreatic cancers. In risk-adjusted models, male sex (0.21[0.06–0.82]; P=0.02), advanced cancer stage (2.29[1.40–3.72]; P<0.01), and surgical resection (0.06[0.01–0.59]; P=0.02) predicted DSS. Median survival time in years increased for males (1.39), earlier stage (0.60), and surgical intervention (5.34).

Limitations

Absent or incorrect reporting in retrospective SEER data is possible. The database is more likely to include academic centers. Some sub-analyses may be underpowered due to limited sample size of a rare cancer.

Conclusions

Our study presents an in-depth assessment of factors that identify high-risk patients. Non-metro residency, Black race, female sex, and no surgical resection each associated with poorer DSS.

Introduction

Granular cell tumors (malignant and non-malignant) account for approximately 0.5% of all soft tissue tumors and often present as asymptomatic, slow-growing, solitary lesions that can be multifocal.¹ They were first described by Abrikossof in 1926 and named myoblastoma due to the histopathologic appearance similar to skeletal muscle cells.² Ultrastructural analysis revealed Schwannian features. For the past half century, they were considered rare neoplasms of neural origin³. Recent reports of non-neural granular cell tumors in the literature have introduced another distinctive entity that shows no obvious lines of differentiation and is not well-characterized^4–7.

The literature addressing the differences in the two subtypes is extremely limited. One study highlights that non-neural granular cell tumors present in a younger age group and with sharper circumscription than those of neural origin⁴. Additional differences in behavior and clinical features have not been well described. It has been suggested that these may not be distinct entities at all but rather a reflection of the heterogeneity of granular cell tumors⁵. Our understanding of the histopathological nature of granular cell tumors is evolving.

This rare tumor is most often located in the dermis or subcutaneous tissues⁸. It is expected to become malignant in less than 2.0% of cases⁹. Local recurrence and metastasis of malignant granular cell tumors when they do occur, however, are thought to be common and associated with a poor prognosis. Malignant tumors have historically been reported to be more likely to develop in the lower extremity versus the head and neck, sites more frequently associated with benign lesions of similar origin^10,11. The most common sites of metastasis in malignant granular cell cases are reported to include the regional lymph nodes, lungs, and bones^10,11.

Although the literature includes a number of case series and reports about granular cell tumors as a whole^12–16, population-level data on malignant cutaneous lesions are limited. The largest study on cutaneous granular cell tumors that included malignant cases analyzed 34 patients with granular cell tumors; only 1 was malignant¹⁷. Clinical presentation and histopathologic analysis of malignant cases has been well profiled for small sample sizes^11,18,19. However, the understanding of patient characteristics and treatment modalities is lacking due to limited data. To that end, the goal of this study was to determine the incidence and survival of patients with all primary malignant cutaneous granular cell tumors – of both of neural and non-neural origin – in a larger population. We used data from the population-based United States (US) National Cancer Institute’s Surveillance, Epidemiology and End Results (SEER) cancer registry to analyze patient and disease characteristics of patients with malignant cutaneous granular cell tumors and to determine factors affecting both overall survival (OS) and disease-specific survival (DSS) after diagnosis.

Methods

Data on patients with a diagnosis of malignant cutaneous granular cell tumor were obtained from the SEER program for the years 1973 to 2013. The database is a widely-used cancer registry that covers an estimated 27.8% of the US population. Geographic regions covered include: San Francisco-Oakland, metropolitan Detroit, Seattle (Puget Sound), metropolitan Atlanta, San Jose-Monterey, Los Angeles, greater California, Connecticut, Hawaii, Iowa, New Mexico, Utah, Kentucky, New Jersey, and rural and greater Georgia, Kentucky, and Alaska. SEER contains no personal identifying information; therefore, this study was deemed exempt from full-review by the Yale Human Investigation Committee.

Patients were identified using the histology code 9580 (malignant granular cell tumor) and primary site codes for the skin (C44.0–C44.9) or for subcutaneous, connective, and soft tissue (C49.0–C49.9). OS was defined as the time in months from diagnosis to death from any cause; DSS as the time from diagnosis to death specific to the cancer-related diagnosis. The SEER cause-specific death classification field was used for DSS, a variable that uses algorithms to process a single, disease-specific underlying cause of death from death certificates. Though there is a risk of misattribution, this is largely mitigated by the coding scheme for this variable which accounts for tumor sequence, site of original cancer diagnosis, and comorbidities. Therefore, our results should provide reliable estimates of death due to granular cell tumor malignancy specifically. Year of diagnosis, continuous age at diagnosis, sex, race, residency demographic, tumor extent and size, body site, treatment with surgery and/or radiation therapy, and OS and DSS in months were extracted from the database for analysis.

Race was categorized as White and Black; all other groups combined made <5.0% of the cohort. Residency demographics were defined using US Department of Agriculture Rural-Urban Continuum codes²⁰. The 9-level scale codes gradations of metropolitan areas (i.e. a population of 20,000 or more) (codes 1–3) and non-metropolitan areas (i.e a population of less than 20,000) (4–9 or Alaska and Hawaii)²¹, thus allowing for the extraction of a dichotomous metro-nonmetro variable. Tumor (T) stage, lymph node involvement (N), and metastasis (M) was retroactively determined where possible using extent of disease and collaborative stage staging codes for tumor size and location, number of positive lymph nodes, and metastatic flags, as well as lymph node and metastasis data. Stages were determined following the 7^th edition of the American Joint Committee on Cancer soft tissue tumor classification protocol²². As data quality made it impossible to distinguish between stage II and stage III lesions, the SEER staging criteria groupings were applied as follows: (1) local – neoplasm confined entirely to organ of origin, or stage I; (2) regional – neoplasm with surrounding extension and/or regional lymph node involvement, or stage II or stage III; and (3) distant – evidence of metastasis, or stage IV.

Descriptive statistics were calculated for all variables. Cell counts less than 5 were redacted to protect patient confidentiality. The OS and DSS curves were calculated using the Kaplan-Meier method, and differences were formally tested for univariate analyses using log-rank tests. A risk-adjusted multivariable Cox proportional hazards regression model was used to assess the independent relationship of reported disease and patient characteristics with OS and DSS. Patient fields with missing data were recoded as empty cells to avoid inaccurate regressions on ordinal variables. Threshold for significance was based on a two-sided P ≤ 0.05. Significant univariate variables were included as covariates along with age, sex, and year. In order to quantify the degree of survival difference, risk-adjusted fiftieth-percentile (median) quantile regression for DSS was used. All statistical analyses were conducted using SAS University Edition. Statistically significant increased risk of secondary malignancies was determined using the Multiple Primary Standardized Incidence Ratio function in SEER*Stat. This report adheres to The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelinesⁱ.

Results

The SEER database search yielded 113 patients with granular cell tumors 1973 to 2013, of which 82 patients (72.6%) presented with tumors on the cutis or subcutis. Of the selected cohort, 65.0% were female and 70.8% were White. The mean age of diagnosis was 49.2 years. At presentation, 15.9% of tumors were metastatic (stage IV), and 17.1% demonstrated regional involvement (stage II/III). The most frequent body site of the lesion was the trunk at 32.5% of the cohort. Surgery was performed in 85.0% of patients, and radiation therapy was given to 11.3% of patients. Patients diagnosed with a malignant cutaneous granular cell tumor are at an increased risk (P<0.05) for subsequent diagnosis of either renal (93.03 per 10,000 person-years [PY]) or pancreatic cancer (89.56 per 10,000 PY) between 60 to 119 months compared to the general population. They were also at increased risk of pancreatic cancer (16.41 per 10,000 PY) during the entire lifetime of the patient (table 1). Patients developed either renal or pancreatic cancer but not both3. The mean time to diagnosis of a second tumor was 113 months (table 1).

Table 1.

Characteristics of patients with malignant cutaneous granular cell tumors.

Characteristic		Valuea
Total		n = 82
Year of diagnosis, y
Mean (SD)		1998 (12.3)
Median (range)		2001 (1973–2013)
Age, y
Mean (SD)		49.2 (18.0)
Median (range)		52 (9–87)
Sex, %
Male		35.0
Female		65.0
Race, %
White		70.8
Black		25.0
Residency demographic
Metro		93.9
Non-metro		6.1
Stage at presentation, %
Stage I (Local)		48.8
Stage II/III (Regional)		17.1
Stage IV (Distant)		15.9
Body site, %
Face		11.3
Trunk		32.5
Upper Limbs		21.3
Lower Limbs		25.0
Pelvis		8.8
Surgery, %
Yes		85.0
No		15.0
Radiation therapy, %
Yes		11.3
No		86.3
Significant increased riskb,c
60–119 months	Pancreas	89.56
	Kidney	93.03
Overall	Pancreas	16.41
Mean time to secondary cancer diagnosis
Overall (months)		113

^aPercentages may not total 100% because of rounding and missing indicator variables for unknown variables (not included in table).

^bIncreased Multiple Primary Standardized Incidence Ratio (MP-SIR) per 10,000 persons per year

^cP<0.05

Survival analysis from Kaplan-Meier curves revealed that the 5-year OS and DSS for malignant cutaneous granular cell tumors was 63.3% and 62.8%, respectively. DSS for patients diagnosed with a local (stage I) tumor was roughly 20 years, a regional (stage II/III) tumor <5 years, and a distant (stage IV) tumor <0.5 years (table 2). The median OS and DSS was 11.6 years and 20.1 years, respectively (figure 1).

Table 2.

Survival data by tumor stage.

Survival	Overall	Disease-Specific
Median (SE)a, y	11.6 (2.0)	20.1 (1.2)
Stage I (Local)	21.5 (2.8)	21.5a
Stage II/III (Regional)	5.3a	3.6a
Stage IV (Distant)	0.6 (3.0)	0.4 (0.5)
Proportion, %
At 1 y	87.2	84.6
At 5 y	63.3	62.8
At 10 y	51.1	55.5

Abbreviation: SE, standard error.

^aUnable to calculate SE of median using a 95% CI for this sample.

Figure 1.

Malignant cutaneous granular cell tumors. Survival analysis using Kaplan-Meier analysis.

A. Kaplan-Meier estimates of overall survival (OS) for all cases.

B. Kaplan-Meier estimates of disease-specific survival (DSS) for all cases.

C. Kaplan-Meier analysis of OS stratified by stage.

D. Kaplan-Meier analysis of DSS stratified by stage.

Both OS (hazard ratio HR[95%CI]: 1.84 (1.40–2.41)) and DSS (2.08 [1.56–2.77]) showed a difference in survival based on stage (P<0.01). Additional univariate survival analyses on patient characteristics revealed that Black race (1.91 [1.00–3.70]; P=0.04) was associated with significantly shorter OS and metro residency (0.28 [0.10–0.80]; P = 0.05) was associated with significantly longer OS. Advanced ageⁱⁱ (1.04 [1.02–1.06] and 1.03 [1.01–1.06]; P<0.01), a lesion on the pelvis versus other body regions (categorical group P<0.01), receipt of surgical resection (0.32 [0.16–0.67] and 0.23 [0.11–0.48]; P<0.01), and no radiation (0.25 [0.11–0.56] and 0.26 [0.12–0.58]; P<0.01) were associated with improved OS and DSS (table 3).

Table 3.

Univariate analysis of overall and disease-specific survival.

		Overall		Disease-Specific
Characteristic		HR (95% CI)	P Value for Log Rank	HR (95% CI)	P Value for Log Rank
Overalla
Year of diagnosis (Advanced)		1.00 (0.98–1.03)	0.98	1.02 (0.99–1.05)	0.43
Age (Advanced)		1.04 (1.02–1.06)	<0.01	1.03 (1.01–1.06)	<0.01
Sex (Male)		0.82 (0.42–1.57)	0.36	0.50 (0.22–1.10)	0.06
Race (Black)		1.91 (1.00–3.70)	0.04	1.23 (0.57–2.65)	0.56
Residency demographic (Metro)		0.28 (0.10–0.80)	0.05	0.45 (0.16–1.27)	0.12
Stage (Advanced)		1.84 (1.40–2.41)	<0.01	2.08 (1.56–2.77)	<0.01
Body site (Pelvis)		1.00 (Reference)	<0.01	1.00 (Reference)	<0.01
Face		0.24 (0.07–0.85)		0.14 (0.03–0.70)
Trunk		0.26 (0.10–0.73)		0.32 (0.12–0.88)
Upper Limb		0.25 (0.08–0.76)		0.20 (0.06–0.66)
Lower Limb		0.33 (0.12–0.94)		0.36 (0.13–1.03)
Surgery (Performed)		0.32 (0.16–0.67)	<0.01	0.23 (0.11–0.48)	<0.01
Radiation therapy (None)		0.25 (0.11–0.56)	<0.01	0.26 (0.12–0.58)	<0.01

On risk-adjusted multivariable analysis, advanced age (HR [95%CI]: 1.05 [1.02–1.08]; P<0.01), advanced stage (2.08 [1.23–3.53]; P<0.01), and lack of radiation therapy (0.26 [0.09–0.79]; P=0.02) were found to be independent predictors of OS. Despite statistical significance, adverse mortality outcomes associated with receipt of radiation therapy are expected to be reflective of more advanced or disseminated disease versus any implicit harm to patients. Advanced age (1.05 [1.02–1.09]; P<0.01), male sex (0.21 [0.06–0.82]; P=0.02), advanced cancer stage (2.29 [1.40–3.72]; P<0.01), and receipt of surgical resection (0.06 [0.01–0.59]; P=0.02) independently predicted DSS. When risk-adjusting for all other covariates, a patient’s DSS: (1) decreased by 0.68 years ([95% CI] −0.22, −0.02) for each additional year of life lived; (2) increased by 1.39 years due to male sex; (3) decreased by 0.60 years (−2.15, 0.00) for each advancing stage; and (4) increased by 5.34 years (3.39, 28.52) through receipt of surgery (supplemental table).

Discussion

Our findings support the previously reported high mortality associated with malignant cutaneous granular cell tumors^11,18, with a 5-year DSS of 62.8%. This study also reveals a number of prognostic factors, of which the most important is thought to be cancer staging. Although there is no clear staging scheme for these tumors, the stratification employed here may be clinically useful for physicians in relaying a prognosis to patients with malignant cutaneous granular cell tumors. The divisions introduced (i.e. local, regional, and distant) are consistent with SEER classifications and, perhaps most importantly, are correlated with DSS in both univariate and multivariable model. The profound differences in median DSS time in years – from 21.5, 3.6, to 0.4 for local, regional, to distant tumors, respectively – is striking and clinically applicable. Analyses in SEER*Stat of the multiple primary standardized incidence ratio also revealed an increased risk of subsequent pancreatic or renal cancers among these patients. Clinically, this finding is important in counseling patients on future risks and has implications for post-diagnosis screening and monitoring. Combined, less than ten renal²³ and pancreatic^24–30 granular cell tumors and have been reported in the literature. Though these cases may contribute to the increased incidence of renal and pancreatic cancers observed in patients with malignant cutaneous granular cell tumors, the biological underpinnings of any such relationship has not been explored. Furthermore, the cases described in this study were coded as true independent renal and pancreatic tumors rather than as metastases or second granular cell primaries. Future work may consider potential molecular and genetic causes of this increased risk.

An analysis of patient characteristics demonstrated demographic, race, and sex-based disparities that have not yet been reported in the literature. These disparities have been demonstrated for a number of other cutaneous malignancies, including but not limited to melanoma³¹, dermatofibrosarcoma protuberans³², and Merkel Cell carcinoma³³. Univariate models demonstrated poorer outcomes for OS amongst patients residing in non-metro areas and those who were Black. Sex differences in DSS in adjusted models were statistically significant, with females suffering worse outcomes; comparatively, their male counterparts lived 1.39 years longer when adjusting for all other covariates for DSS. Females account for the majority of cases, representing 65.0% of this cohort. Importantly, female sex has been found to be related to a prediction of malignancy for intramuscular malignant granular cell tumors³⁴, suggesting a biologic cause of our findings.

Contrary to the current understanding^35–37, our results indicate that radiation therapy – adjuvant or otherwise – is not associated with improvements in OS or DSS based on our risk-adjusted models. It is important to note that the finding that lack of radiation therapy is associated with improved risk-adjusted OS is likely to be due to an underlying clinically-derived selection bias among patients who undergo radiation. They are more likely to include those who are too sick to tolerate surgery or who present with more advanced disease not amenable to surgical resection. Overall, radiation and nonsurgical therapy does not correlate with improved survival in patients with malignant cutaneous granular cell tumors. However, the multivariable analyses revealed that surgical management of disease improves DSS by 5.34 years (3.39, 28.52). Supported by our findings, surgical resection with wide local excision has been repeatedly reported as the standard of care^38–40.

There are limitations in this study design. There is a potential for absent or incorrect reporting in retrospective data, variations in coding practices over time, migration of patients in and out of SEER registry areas, and varied reports on the representativeness of the database itself (i.e. not all hospitals participate, and those that do are more likely to be large academic centers^41–43). Despite these limitations, use of SEER to consider determinants of rare cutaneous malignancies has been validated in the literature^32,44–46. Given the rare nature of the tumor, limited sample size means that significant results are likely to be even more pronounced. The geographic and socioeconomic diversity of the SEER registries make our findings likely to be generalizable across care settings and patient populations in the US.

Conclusions

Herein, we report for the characteristic epidemiology of malignant cutaneous granular cell tumors using available population-level data. Determinants of survival include age at diagnosis, sex, stage, and surgical resection, and there are apparent demographic, race, and sex-based disparities in survival. Patients in this population also are at an increased risk for pancreatic or renal cancers. These data underlie the importance of early diagnosis and surgical treatment of malignant cutaneous granular cell tumors for optimal outcomes, and outline characteristics that may be used to inform future guidelines to identify and manage high-risk patients.

Table 4.

Cox proportional hazard ratio (HR) for overall and disease-specific survival.

	Overall		Disease-Specific
Characteristic	HR (95% CI)	P Value	HR (95% CI)	P Value
Overalla
Year of diagnosis (Advanced)	0.97 (0.92–1.02)	0.29	1.02 (0.96–1.08)	0.50
Age (Advanced)	1.05 (1.02–1.08)	<0.01	1.05 (1.02–1.09)	<0.01
Sex (Male)	0.46 (0.15–1.41)	0.18	0.21(0.06–0.82)	0.02
Race (Black)	2.45 (0.92–6.52)	0.07	1.36 (0.45–4.12)	0.59
Residency demographic*	--	0.26	0.47 (0.04–6.10)	0.57
Stage (Advanced)	2.08 (1.23–3.53)	<0.01	2.29 (1.40–3.72)	<0.01
Body site*	--	0.64	--	0.17
Surgery (Performed)	0.24 (0.03–1.89)	0.17	0.06 (0.01–0.59)	0.02
Radiation therapy (None)	0.26 (0.09–0.79)	0.02	0.34 (0.11–1.03)	0.06

^aCategory in parentheses defines the strata the hazard ratio represents.

^*Residency demographic and body site was included because it was significant on univariate analysis. Inclusion in the risk-adjusted model resulted in insufficient power to make the individual categorical hazard ratios meaningful.

Capsule Summary.

• Knowledge of malignant cutaneous granular cell tumor epidemiology is extremely limited.

• Patients are at an increased risk for pancreatic and renal cancers. Younger age, male sex, earlier stage, and surgical intervention are independent predictors of improved survival.

• These results may facilitate identifying and managing high-risk patients, and prioritizing surgical treatment.

Abbreviations

DSS

Disease-specific survival

HR

Hazard ratio

mcGCT

Malignant cutaneous granular cell tumors

OS

Overall survival

PY

Person-years

STROBE

Strengthening the Reporting of Observational Studies in Epidemiology

SEER

Surveillance, Epidemiology and End Results

US

United States