or less commonly with radiation (7Á9%) or chemotherapy (2Á3%). Wide local excision was the most common surgical technique used (38Á9%). The 5-and 10-year OS rates were 68Á8% and 54Á3%, respectively, with a median follow-up time of 4Á7 years (range 0-13Á8). On Cox proportional hazards regression, older age [hazard ratio (HR) 1Á08, 95% confidence interval (CI) 1Á06-1Á10]; tumours 20-40 mm (HR 2Á37, 95% CI 1Á15-4Á88), 40-60 mm (HR 4Á43, 95% CI 1Á68-11Á7) and > 60 mm (HR 4Á74, 95% CI 1Á63-13Á8); and stage IV disease (HR 5Á95, 95% CI 1Á84-19Á3) were independently associated with decreased OS. Lesions of the lower extremities and hip were associated with a more favourable prognosis (HR 0Á46, 95% CI 0Á25-0Á84). The tumour size varied greatly in our study cohort, with a mean diameter of 23Á88 mm and a range in size of 1-130 mm. Larger tumour size was strongly and independently associated with decreased OS. This finding is consistent with the AJCC staging guidelines for nonmelanoma and non-Merkel cell skin cancers, including EPC, in which tumour diameter > 2 cm is associated with worse prognosis. 3 Other studies of EPC have shown a similarly wide range in tumour size, but with no definitive relation to prognosis. 4 Our study found that 8Á1% of patients had metastatic disease at diagnosis. Patients with distant metastases (AJCC stage IV) had a poor prognosis, consistent with the literature. 4-7 However, in our study, metastases at diagnosis to lymph nodes and/ or distant organs were less common than previously reported in some reviews and meta-analyses, which reported rates of metastases at presentation as high as 31%. 5,7 These reviews detailed cases that were locally aggressive or had lymph node or distant metastases. Other studies similarly suggest that aggressive behaviour and metastases may be the exception. 4,6,8 There are several limitations in retrospective, registry-based studies. The NCDB is a clinician-reported database that relies on accurate and complete reporting by contributing institutions. Missing or erroneous data can lead to bias and data misinterpretation. Additionally, the NCDB does not report on treatment-specific outcomes or disease-specific survival, which may overestimate the mortality risk from EPC, especially if patients have comorbidities. EPC is a rare cutaneous tumour of unknown aetiology that has the potential for extracutaneous spread and increased mortality. Thus, it is crucial to diagnose and treat EPC appropriately. Our findings add to the growing body of evidence that early diagnosis and surgical treatment are the main good prognostic factors.