Transcriptome-based classification of resected pancreatic ductal adenocarcinoma enhances prognostic modelling accuracy of overall survival following adjuvant treatment

In pancreatic ductal adenocarcinoma, patient outcomes after resection remain highly variable. Prognostic models are often inaccurate. Our study aimed to improve survival prediction by adding transcriptome-based classification to a validated prognostic model and applying it on a multicenter real-world cohort. Fresh-frozen tumor resection materials were obtained from the Dutch Pancreas Parel biobank. RNA was sequenced if tumor cellularity was >30%. The samples were classified using transcriptome-based classification. Survival differences between transcriptome-based subtypes were studied in patients treated with and without adjuvant chemotherapy. Both patients receiving neoadjuvant treatment (NAT) and patients undergoing immediate surgery were included. Samples of 461 patients were collected, of which 118 samples underwent RNA sequencing. Of those, 39.0% were of the basal-like subtype and 61.0% were of the classical subtype. The basal-like subtype became dominant after NAT (63.3%, p = 0.004). Adjuvant gemcitabine-treated patients with a classical tumor survived longer than those with a basal-like tumor (median overall survival (OS): 22.8 vs. 11.4 months; p < 0.001). In multivariable Cox regression, the classical subtype significantly associated with increased survival (hazard ratio = 0.38; p = 0.002) and adding transcriptome-based subtyping significantly improved the prognostic model (p = 0.002). Subtype and adjuvant treatment independently statistically significantly associated with OS. Transcriptome-based subtyping significantly adds to clinical variables in survival prediction after surgery. The independent associations for subtype and adjuvant treatment with OS indicate that subtypes are prognostic, but not predictive for OS with adjuvant treatment. The provided prognostic information could potentially support treatment decisions and serve as stratification factor.