Abstract Background: Pancreatic ductal adenocarcinoma (PDAC) is characterized by hotspot mutations in the KRAS gene (codons 12, 13 or 61) in 85-90% of cases. Codon 12 KRAS mutations have been detected in pancreatic juice, blood and stool samples from pancreatic cancer cases and represent promising biomarkers for early detection. However, the proportion of tumor-derived KRAS mutations in cell-free DNA fragments (cfDNA) has shown large variations, probably because of the heterogeneity in biosamples and assays tested. Deep sequencing technologies (NGS) allow the identification of low-abundance somatic variants, but have not previously been applied to the detection of KRAS hotspot mutations in cfDNA of PDAC cases. Moreover, variant calling methods have rarely been tested against cancer-free individuals so the proportion of false positives is unknown. We investigated whether deep sequencing of KRAS mutations at codons 12, 13 and 61 in plasma samples could represent a robust assay to distinguish pancreatic cancer from chronic pancreatitis and healthy controls. Methods: We developed an Ion Torrent-based NGS KRAS assay (partial exons 2 and 3, totalling 208bp) to screen cfDNA from plasma samples of 461 PDAC cases, 154 individuals with chronic pancreatitis and 421 healthy controls. cfDNA extraction (>4ng) and sequencing (>1000x coverage on average, and absence of systematic high sequencing error rates on the 208bp) performed well on 431 (93%) PDAC cases, 138 (90%) chronic pancreatitis, and 388 (95%) controls. We fit a robust negative-binomial regression to estimate the distribution of the sequencing errors at each DNA bp position and identified outlying samples, which were considered as KRAS positive when q-value<10-3. We also estimated the detection threshold of our assay using serial dilutions of DNA from SW480 KRAS mutated cell-line (p.G12V) in wild-type DNA. Results: Sequencing of the serial dilutions of KRAS p.G12V mutated DNA indicated a detection threshold at a minor allele frequency of 0.2%. KRAS mutations in cfDNA were detected in 83 (19.3%) PDAC cases (73 on codon 12; 8 on codon 61; 1 on codon 13; and 1 multiple codons, i.e., similar in proportions as reported in tumor tissue from the International Cancer Genomic Consortium); 3 (2.2%) chronic pancreatitis (all on codon 12); and 8 (2.1%) healthy controls (4 on codon 12 and 4 on codon 61). Stage was significantly associated with the proportion of detected mutations in cancer cases (chi-squared p = 0.0005): the proportions of cases with detectable KRAS mutations in plasma were 7.9%, 14.9%, and 31.1% for local, regional, and advanced stages, respectively. Conclusions: The NGS-based KRAS mutation screening is a sensitive approach to detect low allelic fraction in plasma cfDNA, although its utility for early detection is still limited. However, it has the capacity to identify specific KRAS mutations, which could be useful in a panel of other non-invasive biomarkers. Citation Format: Florence Le Calvez-Kelm, Matthieu Foll, Magdalena B. Wozniak, Geoffroy Durand, Priscilia Chopard, Maroulio Pertesi, Tiffany Delhomme, Ivana Holcatova, Lenka Foretova, Vladimir Janout, Eleonora Fabianova, Maxime P. Vallée, Paul Brennan, James D. McKay, Graham Byrnes, Ghislaine Scélo. NGS-based detection of KRAS hotspot mutations in plasma cell-free DNA of pancreatic cancer cases. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 3137.