Statistical Analysis

All statistical analyses were performed using SPSS, version 25.0 (IBM Inc., Armonk, NY). Overall survival was defined as the time from surgery until any-cause death (date of census: January 22, 2020). The continuous Ki-67 scores were categorized based on the cutoffs used for the current World Health Organization (WHO) classification of neuroendocrine neoplasms27,28 into low (<3%), intermediate (3% to 20%) and high (>20%) groups. To determine the most prognostically significant cutoff, the continuous mitotic count variable was dichotomized first at a cutoff of 2 mitoses per 2 mm² (identical to that used to indicate a grade 1 low-grade neuroendocrine tumor in the gastrointestinal tract) and then at a cutoff of 3 mitotic figures per 2 mm². The other continuous variables were the proportion of amyloid/fibrosis and the proportion of tumor showing spindled morphology. These were each arbitrarily dichotomized at a cutoff of 50%. Kaplan-Meier curves were constructed for each morphologic parameter. Univariate Cox proportional hazard regression analyses were used to evaluate the association between age, sex, MEN2 syndrome, tumor size, vascular space invasion, lymph node metastasis, Ki-67 proliferative index, mitotic count, coagulative necrosis, spindled morphology, sheet-like growth, incipient necrosis, nuclear grade, multinucleated cells, prominent nucleoli, and fibrosis/amyloid deposition and overall survival. A multivariate analysis was performed using the variables that were found to be significant in univariate analyses. Serum calcitonin doubling time was calculated using the American Thyroid Association online calculator (available at: www.thyroid.org/professionals/calculators/thyroid-cancer-carcinoma/). For the purposes of survival analyses, the continuous calcitonin doubling time variable was dichotomized into long and short categories using a cutoff of the median value. The first calcitonin level was also categorized as high or low using a cutoff at the median value. Kaplan-Meier survival analyses were performed using the categorical calcitonin doubling time and the first serum calcitonin variables. All statistical tests were 2 sided and P-values ≤0.05 were considered significant.