LAUSR

Background Pancreatic colloid carcinoma (CC) is a subtype of pancreatic ductal adenocarcinoma (DAC) with low incidence but high malignancy. Unfortunately, there is no consensus regarding the clinical features and prognostic factors associated with CC, and the prognosis is unpredictable. We aimed to assess the clinicopathological characteristics of this rare disease and develop a nomogram for predicting cancer-specific survival (CSS) in CC. Methods We gathered comprehensive clinicopathological data from the Surveillance, Epidemiology, and End Results (SEER) database on 17,617 patients with DAC and 561 individuals with CC. Kaplan-Meier was used to plot each survival curve. Subsequently, we split the 561 patients with CC in a 7:3 split ratio between an internal training cohort (n=393) and an external validation cohort (n=168). The independent prognostic factors for CC patients in the training cohort were discovered using univariate and multivariate Cox regression analyses, and a nomogram was created. We assessed the nomogram’s performance by using the concordance index (C-index), the area under the receiver operating characteristic curve (AUC), calibration curves, and decision curve analysis (DCA). Results The median for follow-up of CC patients was 15 months (range: 1–163 months), and the 1-, 3-, and 5-year CSS were 58.4%, 30.2% and 22.6%. For CC patients in the training cohort, age [hazard ratio (HR) =1.29; 95% confidence interval (CI): 1.00–1.65], sex (HR =0.64; 95% CI: 0.51–0.81), T3 stage (HR =2.21; 95% CI: 1.26–3.88), T4 stage (HR =2.76; 95% CI: 1.47–5.18), N1 stage (HR =1.29; 95% CI: 1.02–1.63), M1 stage (HR =1.60; 95% CI: 1.17–2.18), surgery (HR =0.30; 95% CI: 0.22–0.42), and radiotherapy (HR =0.76; 95% CI: 0.58–1.01) were the main predictors of the nomogram. The C-indexes of the training cohort and the validation cohort were 0.734 and 0.732, respectively. The 1-, 3-, and 5-year AUC values of the nomogram were predicted to be 0.827, 0.816, and 0.831 in the training cohort, 0.801, 0.841, and 0.835 in the validation cohort, respectively. Conclusions Based on several clinical features, we established the first predictive model of CC. This nomogram could be used to guide treatment decisions in patients with CC.