AASLD Abstracts of nucleoside analogues was analyzed as a time dependent covariate. Results: On modified Cox proportional hazards analysis, nucleoside analogue (HR 0.75, 95% confidence interval [CI], 0.65-0.87, P,0.001) was independently associated with a reduced risk of death, after adjusting for age, gender, presence of liver cirrhosis, liver decompensation, vascular invasion of tumor, diabetes, and use of statin and metformin. Multivariable stratified analyses further verified the beneficial effect of nucleoside analogues in several subgroups of patients, including those younger than 50 years (HR 0.57, 95% CI 0.42-0.77), male gender (HR 0.73, 95% CI 0.62-0.85), with cirrhosis (HR 0.74, 95%CI 0.62-0.88), without statin use (HR 0.75, 95% CI 0.65-0.87) and without metformin use (HR 0.72, 95% CI 0.62-0.85). The status of liver decompensation, vascular invasion of tumor, or diabetes did not affect the beneficial effect of nucleoside analogues on survival among study patients. Conclusions: Nucleoside analogue therapy is associated with a lower risk of death among patients with HBV-related HCC after TACE. 958 Patient-Level Determinants of Alpha Fetoprotein Elevation Among Patients With Cirrhosis Amit G. Singal, Jason Chiang, Purva Gopal, Mahendra S. Nehra, Pragathi Kandunoori, Adam Yopp Background: The utility of alpha fetoprotein (AFP) as a surveillance tool for hepatocellular carcinoma (HCC) is limited by insufficient sensitivity and specificity. However, few studies have attempted to identify subsets of patients in whom AFP may be more accurate for the presence of HCC. Aims: To identify patient-level determinants of elevated AFP levels among patients with cirrhosis. Methods: We conducted a case-control study of cirrhotic patients, with and without HCC, seen at Parkland Hospital between January 2005 and July 2012. Demographic, lab, and clinical data were abstracted using standardized forms. HCC diagnosis was confirmed using AASLD guidelines and staging done using the Barcelona Clinic Liver Cancer (BCLC) system. The maximum AFP level over a one-year period was recorded for non-HCC patients and compared to the AFP level at diagnosis in those with HCC. The accuracy of AFP was assessed by receiver operator characteristic (ROC) curve analysis and compared between groups using the Delong method. Results: We identified 1100 patients with cirrhosis (457 with HCC and 643 non-HCC patients). The cohort was diverse with respect to race (38% Hispanic, 33% Caucasian, and 25% African American), 73% were male, and 64% had viral hepatitis (58% HCV and 6% HBV). Patients were diverse with respect to tumor stage (29% BCLC stage A, 9 BCLC B, 29% BCLC C, and 33% BCLC stage D). The median AFP level for non-HCC patients was 5 ng/mL (range 1-1105), with 11.7% having a level greater than 20 ng/mL. Patients with HCC had a median AFP level of 195 ng/mL, with 70.2% having a level greater than 20 ng/mL. AFP .20 ng/mL was accurate for presence of HCC, with a c-statistic of 0.81 (95%CI 0.78-0.83). An elevated AFP better predicted presence of HCC among patients without hepatitis C virus (HCV) infection (p= 0.007), those with platelet count .100 *109/L (p=0.02), and those with AST .40 U/L (p= 0.05). Among HCV-positive patients, AFP .20 ng/mL had a c-statistic of 0.83 (95%CI 0.79 - 0.87), compared to 0.76 (95%CI 0.72 - 0.79) among patients without HCV. Although the sensitivity of AFP for the presence of HCC did not differ between HCV and non-HCV patients (70.4% and 69.8% respectively), its specificity was significantly higher in non-HCV patients (96.3% vs. 81.1%). Conclusions: AFP has limited specificity for HCC in HCV- positive patients; however, AFP levels greater than 20 ng/mL are likely sufficiently accurate in non-HCV patients to be beneficial during HCC surveillance. 959 Early Prediction of Response to Sorafenib in Patients With Advanced Hepatocellular Carcinoma: the Role of Dynamic Contrast Enhanced Ultrasound Maria Assunta Zocco, Andrea Lupascu, Enrico Di Stasio, Maria Elena Ainora, Matteo Garcovich, Davide Roccarina, Brigida E. Annicchiarico, Gianluigi Caracciolo, Francesca Romana Ponziani, Francesca D'Aversa, Annalisa Tortora, Laura Riccardi, Massimo Siciliano, Gian Ludovico Rapaccini, Maurizio Pompili, Antonio Gasbarrini Background. Sorafenib has become the standard first-line treatment for patients with advanced hepatocellular carcinoma (HCC). This molecularly targeted therapy acts not only by inhibiting proliferation but also by inducing alterations in tumor vascularity that cannot be evaluated by traditional morphological criteria to assess tumor response. Contrast enhanced ultrasound (CEUS) is now recognized as a functional imaging technique able to evaluate tumor vascularization. In particular, recently developed quantitative approaches able to measure the amount and the time course of contrast uptake have shown great promise in revealing effective tumour response to anti-angiogenic drugs before tumour changes occur. Aim. To evaluate the feasibility of dynamic CEUS (D-CEUS) with perfusion software as a predictor of early tumor response to Sorafenib in patients with advanced HCC and to correlate these functional parameters with clinical efficacy endpoint. Patients and methods. Consecutive patients with advanced stage HCC treated with Sorafenib, 400 mg bid, were prospectively enrolled in this study. In patients with tumor targets that were accessible to US, CEUS with second generation contrast agent (Sonovue, Bracco) was performed at baseline (T0) and after 15 (T1) and 30 (T2) days of treatment. Tumor vasculature was assessed in a specific harmonic mode associated with a perfusion and quantification software (Q-Lab, Philips). Variations between T1/T2 and T0 were calculated for five D-CEUS functional parameters (peak intensity, PI; time to PI, TP; area under the curve, AUC; slope of wash in, Pw; mean transit time, MTT) and were compared for responders and non responders. The correlation between D-CEUS parameters and overall survival (OS) end time to progres- sion (TTP) was also assessed. A p value , 0.01 was considered statistically significant. Results. We report the preliminary results obtained in 20 patients treated with Sorafenib. The percentage variation at T1 significantly correlated with response in three of the five D- CEUS parameters (AUC, PI and Pwi, p = 0.006, 0.008 and 0.009 respectively). No statistically significant correlation was found between D-CEUS parameters and OS or TTP. However a trend toward correlation with TTP was found for TPI (p=0.035) and Pwi (P = 0.014). Changes in tumor vascularisation observed at T1 were always consistent in the same patients at T2. Conclusion. Dynamic US provide a more reliable and early measure of efficacy for S-968 AASLD Abstracts antiangiogenic therapies and could be an excellent tool for selecting patients who will benefit from treatment. However further studies with a greater number of patients are required to confirm these results. 960 Alpha-Fetoprotein Should Not Be Excluded As Part of Surveillance for Hepatocellular Carcinoma in the Real World Setting Maria Carla V. Tablante, Jose D. Sollano, Stephen N. Wong Background: Recent guidelines have stricken out alpha-fetoprotein (AFP) and recommended ultrasound only as screening for hepatocellular carcinoma (HCC) in at-risk patients. However, AFP may still need to be done in the real world where ultrasound quality is very variable. Objectives: We aimed to determine the frequency at which ultrasound misses a diagnosis of HCC in patients with elevated AFP, and to characterize tumor and patient factors that may contribute to the missed diagnosis. Methods: Patients with chronic liver disease and elevated AFP levels done within 12 weeks of a normal ultrasound (no tumor) from January 2006 to November 2012 were included. The diagnosis of HCC was confirmed/excluded either by dynamic imaging studies or biopsy done within 12 weeks of the ultrasound, or by serial ultrasound monitoring of at least 24 weeks duration. Baseline demographics and laboratory data were compared between patients with HCC and with no HCC. Results: Out of 341 patients with elevated AFP, only 41 (12%) had a normal ultrasound and were included in the study. The rest were excluded because of hepatic mass/es on ultrasound. Majority (30/41; 73.2%) of the patients were ruled out to have HCC through dynamic imaging (N= 25) or through long-term (median=125 weeks) follow-up (N=5). However, 2 of these patients eventually developed HCC 66 and 87 weeks after the initial ultrasound. The most common cause of elevated AFP in patients without HCC was cirrhosis (43.3%), followed by hepatic flare (36.7%). The most common AFP pattern on follow-up was a fluctuating pattern (46.7%). Only a minority (13.3%) normalized AFP at the last follow-up. Of the 11 patients diagnosed with HCC despite no tumor on the initial ultrasound, 5 were proven to have HCC on biopsy and all had typical findings of HCC on dynamic imaging. Majority (9/11; 81.8%) had a diagnosis of cirrhosis on ultrasound while only 2 (18.2%) were deemed to have a "normal" ultrasound. Mean tumor size was 8.3 cm (range 3-16) and 4 (36.4%) were infiltrating in morphology. Alarmingly, 36.4% already had tumor thrombus on diagnosis. Patients with HCC were more likely to be male (90.9% vs. 46.7%; p=0.011) and have higher baseline AFP (478.4+455 ng/mL vs. 102.1+248.6 ng/mL;p=0.023) compared to patients without HCC. There was no difference in where the ultrasound was performed (city vs. provinces) Conclusions: More than a quarter of patients with elevated AFP despite a normal ultrasound were found to have HCC on further evaluation. The presence of cirrhosis and an infiltrating tumor type may interfere with the accuracy of ultrasound. This highlights the pitfall of excluding AFP as part of surveillance for HCC in the real world setting. 961 The Presence of Portal Vein Thrombosis Alters the Classic Enhancement Associated With Diagnosis of Hepatocellular Carcinoma Nadia Umar, Marwan Ghabril, Kumar Sandrasagaran, Paul Y. Kwo The diagnosis of hepatocellular cancer (HCC) by axial imaging requires arterial enhancement followed by washout in the portal venous phase for lesions greater than 1 cm. Portal vein thrombosis is a common finding in the setting of cirrhosis. Our AIM was to determine if the presence of portal vein thrombosis (PVT) where venous flow within the liver may be altered may delay the diagnosis of HCC and be associated with more advanced disease. We characterized the incidence and imaging characteristics of patients diagnosed with hepatocellular carcinoma in a cohort of patients with PVT compared to those without PVT. Methods: Single center retrospective study of all patients with TIPS or portal vein thrombosis who were subsequently diagnosed with hepatocellular carcinoma between January 2000 and August 2012. The study included all patients with newly diagnosed HCC by dynamic imaging or confirmed with pathology in cases of atypical enhancement. Data abstracted included demographic data, TNM stage, number/type of scans, AFP level, MELD score, time to diagnosis. Results: 92 patients newly diagnosed with HCC on surveillance were reviewed, of which 37% (30/82) of patients were also found to have portal vein thrombosis. Patients with PVT had greater rates of atypical imaging associated with HCC, with lower rates of portal venous washout compared to those without PVT with HCC (table). Patients with PVT and HCC were also diagnosed at later TNM stage than those without PVT. 14/30 had pre-existing PV clot without HCC, 16/30 developed PVT during surveillance. Those with preexisting PVT were older (63. vs 55 years) and had higher rates of diagnosis of HCC using MRI (78.6% vs 21.4% with CT, p=0.01), compared to those without pre-existing PVT. Conclusion: The presence of PVT found on dynamic imaging was associated with more advanced stage of HCC at time of diagnosis. Clinicians should have a high suspicion for HCC diagnosis in new liver lesions with atypical enhancement in the setting of PVT. In this setting, MRI was more frequently associated with HCC diagnosis.