Diagnostic and predictive factors in pancreatic cancer : clinical and translational studies

Abstract

Aims: The aim of Study I was to assess the effectiveness of chemotherapy for patients with advanced pancreatic cancer in a real-world setting. Study II aimed to evaluate the utility of a clinical decision support system to identify precision oncology opportunities for pancreatic cancer. The aim of Study III was to identify the blood component most enriched with pancreatic cancer-derived circulating DNA; that of Study IV was to characterize long and short tumor-derived circulating DNA fragments.

Methods: Study I is a single-institution retrospective cohort study of prospectively generated clinical data. We included a total of 595 patients and used univariate and multivariate models, including flexible parametric models to analyze overall survival and time to treatment failure according to different first-line chemotherapy regimens. Exploratory analyses included the adherence to different protocols, adverse events, and second-line chemotherapy. Study II is a prospective observational study of 39 patients with pancreatic cancer who were enrolled in the PePaCaKa-01 study. Archival tumor tissue was sequenced, and data was processed with the proprietary clinical decision support system MH Guide and results were evaluated by a study-specific molecular tumor board. Endpoints of the study were the frequency of successful generation of support system reports, the frequency of actionable molecular targets, and their evaluation by the tumor board. We performed a post-hoc analysis to determine the proportion of patients who received molecular informed therapies. Studies III– IV analyzed blood samples from a prospective cohort of patients with advanced pancreatic cancer. We systematically separated whole blood into red and white blood cells, platelets, apoptotic bodies, large and small extracellular vesicles, and soluble protein using differential centrifugation. We confirmed efficient separation with protein assays (Western blotting and multiplex bead-based extracellular vesicle flow cytometry), nanoparticle tracking analysis and transmission electron microscopy, and extracted DNA from all components. We used digital PCR to quantify the abundance of KRASmut DNA, a hallmark of pancreatic cancerderived DNA. In Study IV we additionally used automated electrophoresis to quantify the lengths of circulating DNA fragments and ligation-based sequencing library preparation and tagmentation to selectively target short and long DNA fragments, respectively.

Results: In Study I, we observed similar overall survival for gemcitabine/capecitabine, gemcitabine/nab-paclitaxel, and FOLFIRINOX (including modified regimens) compared to gemcitabine. Combinations of 5-fluorouracil/oxaliplatin and best supportive care were associated with poorer outcomes. Models adjusting for other demographic and clinical variables showed a survival benefit for gemcitabine-combinations and FOLFIRINOX. Exploratory analyses revealed differences in protocol adherence across different treatments, a relatively low frequency of AEs, and a difference between different sequences of first- and second-line therapy. In Study II, a CDSS report was generated for 31/39 patients, 28/31 reports were evaluated at the study-specific molecular tumor board The clinical decision support system made 80 individual recommendations to use molecularly informed therapies based on 61 genomic variants. In 21/28 cases, the tumor board classified at least one molecularly informed therapy as a potential clinical option. The post-hoc analysis revealed that six patients received molecularly informed treatment in routine care. In Study III, KRASmut DNA had the highest concentrations in the soluble protein and small vesicles blood fractions at late stages of PDAC. At early stages, it was highest in large and small extracellular vesicles. Small extracellular vesicles also contained the highest ratio of the concentrations of mutant : wild type KRAS DNA at this stage. In Study IV, blood from PDAC patients had significantly higher concentrations of short cell-free DNA in the soluble protein fraction than that of healthy individuals. The mutant allele frequency of KRASmut was highest in this blood component. Long genomic DNA fragments were most reliably measured in association with apoptotic bodies but KRASmut genomic DNA occurred in all assessed blood fractions.

Conclusions: Chemotherapy in clinical routine use can result in outcomes that reflect relevant randomized controlled trials and gemcitabine-based regimens are highly effective this setting. Differences between different treatments might be related to how they are applied (Study I). The clinical decision support system MH Guide could identify clinically relevant opportunities for molecularly informed treatments of advanced pancreatic cancer (Study II). At early stages of pancreatic cancer, tumor-derived DNA is mostly associated with small extracellular vesicles; in more advanced disease it is mainly a feature of the soluble protein fraction (Study III). Short DNA fragments in this fraction are a more robust source of tumorderived DNA than longer genomic DNA fragments (Study IV).