pancreatic duct (PD), presence of mural nodule or solid mass, unilocular/multilocular, macrocystic or microcystic morphology, and impression of a pseudocyst] were noted. FNA results, including cyst fluid CEA , and cytology features of mucin, cellular atypia or cancer were recorded. Univariate analysis (Fisher's exact and Mann-Whitney U test) was performed followed by multivariate logistic regression analysis to identify if CEA was an independent predictor for MCN. ROC curves were generated for CEA levels to evaluate its diagnostic characteristics. Results: 2407 pts met inclusion criteria and 1861 (77.6%) underwent EUS- FNA [mean age 63 yrs, 87% Caucasians, 58% females, and mean pancreatic cyst size 26 mm]. 376 pts underwent surgery (37% Whipple's surgery, 45% distal pancreatectomy, 18% others) and cancer on cytology was noted in 138 pts. Based on the gold standard, there were 440 pts with MCN (including IPMN) and 882 NMCN. On univariate analysis, the median values of CEA in MCN were significantly higher than that of NMCN (126.9 vs. 21.65, p < 0.001). On multivariate analysis, CEA [OR = 1.3 (1.1 - 1.6), p < 0.01] was a statistically significant predictor for MCN while adjusting for mucin on cytology, presence of solid mass on EUS, mural nodule, and multilocularity. Area under ROC curve was 0.74 (0.70 - 0.77, p<0.01) for CEA (cut off of 92.9 ng/ml - sensitivity, 70% and specificity, 64%) (Figure). The traditional cut off of 192 ng/ml yielded a sensitivity of 59% and specificity of 73% in identifying MCN. Conclusions: This large multicenter study demonstrates that cyst fluid CEA levels have an unacceptable accuracy level in differentiating MCN and NMCN of the pancreas limiting its applicability in clinical practice. Future studies should focus on novel markers in cyst fluid to improve the risk stratification for diagnosis, cancer risk and surveillance of pancreatic cystic neoplasms. Table: AUC, Sensitivity and specificity of CEA in differentiating MCN from NMCN. Figure: Receiver operating characteristics (ROC) curve evaluating the performance of CEA in pancreatic cyst fluid analysis (AUROC = 0.74, p<0.01) Tu1920 The Role of Genetic Variation and C-Reactive Protein (CRP) in IBD Gati Goel, Dalin Li, Alexander N. Levy, Ken D. Nguyen, Xiaofei Yan, Shaohong Yang, Phillip Fleshner, Talin Haritunians, Dermot P. McGovern Background: C-reactive protein (CRP) is biologically active as well as a biomarker associated with disease outcomes in IBD. The aim of this study was to identify the role of genetic variation in CRP in disease behavior and also identify associations with serum CRP level. Methods: The role of CRP SNPs in disease behavior (Montreal classification in CD and the need for colectomy in UC) was assessed in 3965 IBD patients (2412 CD: 1384 UC). CD behavior associations were determined separately for all-Caucasians (Ashkenazi Jewish (AJ) and non-Jewish), AJ alone, and non-Jewish Caucasians. UC associations were determined in all-caucasians. In a subset of 147 patients we tested for association between CRP SNPs and normal serum CRP despite evidence of active disease at endoscopy. We also tested for genetic association with level of CRP using immunochip. Logistic and linear regression, after correction for population structure, were used to test for association. Results: In the Caucasian CD cohort, we observed association between CRP SNPs and: younger onset of disease (rs12057245, p=0.03); disease location, Colonic (L2) (rs3093066, p=0.0008); and disease behavior, internal penetrating disease (B3) rs1446966 (p=0.002). In UC rs2808630, (p= 0.01) and rs2794521,( p=0.02) were associated with need for colectomy. In the AJ population CRP SNPs were associated with: Ileal CD (L1), rs3093066, (p=0.001), rs12057245 (p= 0.008), rs3093077 (p=0.02), rs3093059 (p=0.02); Colonic CD (L2), rs3093066, (p=0.001), rs12057245 (p=0.007), rs3093077 (p=0.03), rs3093059 (p=0.03); Non- stricturing or pene- trating disease (B1), rs1130864 (p=0.02), rs1205 (p=0.04); and penetrating disease (B3), rs1446966 (p=0.04). In the non- Jewish population, CRP SNPs were associated with with: younger age at diagnosis, rs3093077 (p=0.01), and rs3093059 (p=0.02); internal penetrating disease (B3), rs1446966, p=0.009); and need for surgical intervention, rs2794521 (p=0.04) and rs2808630 (p=0.04). The rs1446966 shows consistent association with B3 CD in both AJ and non-Jewish populations. No association between ‘non-flare' of serum CRP and CRP SNPs was seen. CRP level was associated with SNPs tagging: GTF2IRD1 (p=0.00008); NEU3 (p=0.00009); CCR6 (p=0.00023); PTPN22 (p=0.00025); ALPK2 (p=0.00025); ZNF365 (p= 0.00026); TNFAIP3 (p=0.00026); The KIR locus (p=0.00041); and ELMO1 (p=0.00092). Conclusion: CRP polymorphisms are associated with IBD severity and phenotype with independent ‘replication' for the association between B3 CD and rs1446966. A number of S-873 AGA Abstracts immune related genes show association with CRP level including known IBD loci and genes associated with other immune-related conditions (MS - ELMO1 and ALPK2; Celiac - ALPK2; Psoriasis - ALPK2). These findings may help enhance the understanding and use of biomarkers for stratifying IBD patients. Tu1921 RBP-J Gene May Be Associated With Bone Mineral Density of IBD Patients Yoichi Kakuta, Arushi M. deFonseka, Jane F. McGovern, Talin Haritunians, Dalin Li, Dermot P. McGovern Background: Inflammatory Bowel Disease (IBD) patients are at increased risk of osteopenia and osteoporosis, but the complete etiology of low bone mineral density (BMD) in IBD is unclear. It is reported that the diagnosis of low-BMD in parents of pediatric IBD patients predicted low-BMD in the proband suggesting, that hereditary factors contribute to low- BMD in IBD. The aim of our study was to use a genome-wide approach to identify genetic factors that contribute to BMD in IBD. Method: 424 Caucasian IBD patients with BMD data and imputed GWAS data (2,017,629 SNPs) were included. T-scores from 3 locations (Lumber Spine, Left Hip and Right hip) were analyzed as a quantitative trait (n=285) and low-BMD, defined as a T-score of less than -1 in at least one of the 3 locations. (n=424), as a binary trait. Linear and logistic regression after correction for population stratification with the top 4 principal components was used in both analyses. A p-value less than 2.48E-8 was considered as genome-wide significant level and less than 5.0E-5 as showing nominal significance. Results: BMD quantitative trait analyses: One SNP (rs10028140) located upstream of RBP- J gene was significantly associated at genome-wide level with BMD of left hip (p=9.45E- 09), and a borderline association with the same SNP was observed at the right hip (4.86E- 07). Nominal significance was seen at PDLIM5 (p=9.81E-7), PMS2P4 (p=9.89E-7), TTC40 (p=9.01E-7) and TSHZ3 (6.22E-07) showed nominal association with BMD in lumber spine and right hip respectively LBD-analyses: No SNPs showed genome-wide association, but SPRY1 (p=7.90E-06) and TAS2R1(6.06E-06) showed nominal associations with LBD. Conclu- sion: We have demonstrated genome-wide significance with BMD and the RBJP-J locus in addition to a number of loci with nominal significance including SPRY1. Mice deficient for Spry1 exhibit decreased bone mass and SPRY1 polymorphisms are associated with osteopo- rosis in Korean women suggesting that SPRY1 may play an important role in low-BMD that is not specific to IBD. RBP-J is a transcription that limits TNF-induced osteoclastogenesis. Furthermore TNF-induced osteoclastogenesis and bone resorption in vivo are restrained by RBP-J. RBP-J is not associated in BMD-GWAS in general populations, but is associated with Rheumatoid arthritis susceptibility. RBP-J may play an important role in decreasing of BMD specifically in TNF-associated diseases such as IBD and so may be an excellent therapeutic target for low-BMD in IBD. Tu1922 Genetic and Clinical Characterization of 41 Multiple-Affected IBD Families Alessia Settesoldi, Thomas Billiet, Eveline Hoefkens, Vera Ballet, Fred Princen, Steven Lockton, Gert A. Van Assche, Paul J. Rutgeerts, Severine Vermeire, Isabelle Cleynen BACKGROUND AND AIM : Inflammatory bowel disease pathogenesis comprises genetic, environmental and immunological factors. The genetic progress has been outstanding with 163 genetic loci identified. Multiple-affected families may have a high burden of genetic risk variants. We wanted to characterize these families clinically and genetically and to find to what extent the 163 loci are associated with IBD in these families. MATERIAL AND METHODS : We collected 41 families with ≥3 1st-degree members affected by Crohn's disease, ulcerative colitis or IBD unclassified. Clinical data (sex, age at diagnosis, disease location and behavior) were obtained. Affected and unaffected members were genotyped using Immunochip for the 163 known genetic loci. Serum samples were analyzed for the Prometheus serology panel including ASCAA, ASCAG, CBir1, Fla2, FlaX, OmpC, ANCA markers. RESULTS : 29 families were CD only, 10 were mixed CD/UC and 2 were mixed CD/IBD-U. These families comprised 122 CD, 15 UC, 3 IBD-U and 112 healthy relatives. Although females were more affected than males, that was not statistically significant (p= 0.065). The median age at diagnosis was 27 years for CD (IQR 20-35) and 34 for UC (IQR 25-46). Members of a same family showed high degree of concordance (at least 2/3) in age at diagnosis (93%), smoking habit (88%), disease location (71%) and behavior (75%). The difference in the average number of risk alleles between affected and unaffected members was not significant, underlining the importance of environmental factors and possibly not yet identified rare genetic variants. 18 SNPs were nominally significant for the parenTDT test. The most significant was rs3172494 (IP6K2), located in the 3p21 region close to MST1. NOD2 frameshift mutation was also significantly associated with disease in these families. The serology markers seemed to be disease-dependent rather than family-dependent with low intraclass correlation coefficients within families (0.06-0.21). Sensitivity, specificity, PPV and NPV for a diagnostic prediction of IBD vs non-IBD were 85.6%, 75.9%, 85% and 76.8%, respectively. CONCLUSIONS : This is the largest cohort of multiple-affected IBD families studied for genetic association to date. We found a high degree of concordance among CD patients of a same family in clinical features. We found nominal significance for disease association with 18 of the known IBD susceptibility loci, but overall they do not seem to explain much of genetic risk in these families. It would therefore be interesting to perform exome or whole-genome sequencing on some of these families to identify additional risk loci. Serology markers could be useful to diagnose disease rather than to predict it. AGA Abstracts