In univariate logistic regression analysis (Table 3), HBeAg positivity had the highest odds ratio (OR) [OR: 2.920, 95% confidence interval (CI): 1.558–5.371, P = 0.001] and was more than twice that of HBsAg positivity (OR: 1.417, 95% CI: 1.019–1.969, P = 0.038). In the subse- quent multivariate analysis with other significant factors, HBeAg positivity was still the strongest predictor of synCRLM (OR: 2.473, 95% CI: 1.081–5.661, P = 0.032) (table 4), the OR of HBsAg positivity was 1.671 (95% CI: 1.060–2.633, P = 0.027) (table 5). Conclusion: HBeAg positivity is a clinical risk factor for CRLM that can be readily identified and addressed. It is yet unclear if antiviral treatment can decrease the risk of liver metastasis in CRC patients, but future studies with carefully designed prospective trials will be needed to better define this. THU470 Identification and etiology-dependent evaluation of diagnostic algorithms for early detection of hepatocellular carcinoma Magdalena Swiatek-de Lange 1 , Henry Chan 2 , Teerha Piratvisuth 3 , Tawesak Tanwandee 4 , Satawat Thongsawat 5 , Wattana Sukeepaisarnjaroen 6 , Juan Ignacio Esteban Mur 7 , Marta Bes 8 , Bruno Köhler 9 , Marcus-Rene Lisy 1 , Christine Anna Belz 1 , Vinzent Rolny 1 , Ying He 1 , Dave Morgenstern 10 . 1 Roche Diagnostics GmbH, Penzberg, Germany; 2 Chinese University of Hong Kong, Department of Medicine and Therapeutics, Hong Kong, China; 3 Prince of Songkla University, NKC Institute of Gastroenterology and Hepatology, Hat Yai, Thailand; 4 Mahidol University, Division of Gastroenterology Siriraj Hospital, Bangkok, Thailand; 5 Maharaj Nakorn Chiang Mai Hospital, Department of Internal Medicine, Chiang Mai, Thailand; 6 Srinagarind Hospital, Department of Medicine, Khon Kaen, Thailand; 7 Universitat Autónoma de Barcelona, Hospital Vall d’Hebrón, Barcelona, Spain; 8 Banc de Sang i Teixits, Facultativa Laboratori de Seguretat Transfusional, Barcelona, Spain; 9 National Center for Tumor Diseases University Hospital Heidelberg, Department of Medical Oncology, Heidelberg, Germany; 10 Roche Diagnostic Operations, Inc., Indianapolis, United States Email: magdalena.swiatek-de_lange@roche.com Background and Aims: Hepatocellular carcinoma (HCC) has limited treatment options when diagnosed late so early detection is crucial to reduce mortality. Multiple factors associated with cirrhosis, e.g. HBV, HCV or steatohepatitis are involved in the etiology of HCC and have an impact on disease course, diagnosis and therapy. Ultrasound is used for the surveillance of at-risk patients, but has limitations in the setting of NASH and severe liver disease. Alpha-fetoprotein (AFP) improves HCC detection but lacks sensitivity for small tumours. With the aim to improve early diagnosis of HCC in different at-risk populations, we identified biomarker panels with high sensitivity and specificity for early HCC and analysed their performance in context of liver disease etiology. Method: Over 60 biomarkers were evaluated in a panel composed of 308 HCC cases including 125 early stages (BCLC 0 and A) and demographically matched 734 chronic liver disease controls (CLD) from a prospective multi-center study. Univariate and multivariate analysis were performed to distinguish early stage HCC and all HCC from CLD. Performance of the selected panels was compared to GALAD score (Gender, Age, AFP-L3, AFP, DCP). Impact of cirrhosis, HBV, HCV and non-viral etiology on clinical performance was investigated for each biomarker panel. Results: In over 100.000 analyzed combinations AFP and PIVKA-II were the strongest features, but with limited sensitivity toward early stages. Addition of a 3rd biomarker: AFP-L3, IGFBP3, COMP or MMP3 together with gender and age further increased accuracy. Performance of the biomarkers was influenced by disease etiology: AFP showed robust performance in HBV, HCV and NASH/ASH; PIVKA- II had limited performance in ASH, while IGFBP3, MMP3 and COMP were affected by ASH/NASH. AFP-L3 performed well in HCV and NASH/ASH groups. Compared to other scores, GALAD showed superior robustness and clinical performance with AUC of 95% for all- and 90% for early HCC for viral- and non-viral etiologies. Performance in the cirrhosis cohort was 4% lower with AUC of 92% for all HCC and 87% for early HCC. Conclusion: Analysis of over 100.000 biomarker combinations identified panels for early diagnosis of HCC with differential impact of etiology on their performance. None of the candidates showed sufficient performance as single marker and none of the identified scores was superior to GALAD. GALAD score showed robust diagnostic performance in both viral and non-viral etiologies and was only slightly affected by cirrhosis. THU471 Optimisation and qualification of tumour mutational burden (TMB) by targeted next-generation sequencing (TNGS) as a clinically applicable biomarker in hepatocellular carcinoma (HCC) Ching Ngar Wong 1 , Kathy Dominy 2 , Francesco Mauri 1 , Takahiro Kaneko 3 , Persephone Du Parcq 2 , Jamshid Khorasad 2 , Pierluigi Toniutto 4 , Robert D. Goldin 5 , Claudio Avellini 6 , David J. Pinato 1 . 1 Imperial College London, Surgery and Cancer, London, United Kingdom; 2 Hammersmith Hospital, Molecular Pathology Laboratory, London, United Kingdom; 3 Tokyo Medical and Dental University, Medicine, Tokyo, Japan; 4 University of Udine, Medical Area (DAME) Hepatology and Liver Transplantation Unit, Udine, Italy; 5 Imperial College London, Centre of Pathology, London, United Kingdom; 6 Institute of Histopathology, Azienda Ospedaliero-Universitaria “Santa Maria della Misericordia”, Udine, Italy Email: david.pinato@imperial.ac.uk Background and Aims: Higher levels of somatic non-synonymous mutations associate with improved response and survival following immune-checkpoint inhibitors. In this pilot study, we aimed to optimise tNGS as a method to provide a reliable estimate of patients’ TMB in HCC. Method: Following macrodissection and DNA purification, 48 samples derived from an international biorepository (21 fresh- frozen [FF] and 27 formalin-fixed paraffin-embedded, [FFPE]) underwent tNGS by OncomineTM Tumour Mutation Load Assay (1.5 Megabase exome coverage) on an Ion S5TM sequencer. We performed uracil-DNA glycosylase (UDG) pre-treatment in a group of 11 FFPE samples to verify its effect on fixation-induced cytosine deamination. In total, 30/48 samples satisfied post-sequencing quality control and were included for clinicopathological correlation. We classified samples as high/low TMB based on median number of mutations/Mb (Mut/Mb), testing different minimum allele frequency (MAF) thresholds (≥0.05, ≥ 0.1 and ≥ 0.2) in relationship with clinicopathological features including Overall (OS) and Recurrence- Free survival (RFS). Results: Eligible patients (n = 30) were mostly male, cirrhotic (84%) secondary to alcohol (48%) and Hepatitis C infection (45%). Median dominant tumour size was 4 cm with most patients being staged as TNM stage I-II (75%). Median OS was 12.3 months and median RFS was 12.1 months. FFPE samples displayed significantly higher TMB (median 958.39 vs 2.51 Mut/Mb, p < 0.0001), estimated deamination counts (median 1335.50 vs 0, p < 0.0001) and percentage C > T transition at CpG sites (median 60.3% vs 9.1%, p = 0.002) compared to FF. UDG-treated FFPE samples carried significantly lower TMB (median 4019.92 vs 353 Mut/Mb, p = 0.041), estimated deamination counts (median 6393.5 vs 328.5, p = 0.041) compared to untreated FFPE. Adjustment of MAF to 0.1 and 0.2 reduced deamination counts (p < 0.05), with a 0.2 threshold allowing for a reduction of TMB values within the interpretable range of <100 Mut/Mb in all samples. At a 0.2 MAF threshold with UDG treatment, the median number of non-synonymous mutations/Mb was 5.48 (range 1.68– 16.07) and did not correlate with salient pathologic features of HCC including OS or RFS. Conclusion: This pilot study highlights the challenges of TMB testing in archival tissue, where UDG pre-treatment and MAFadjustment to 0.2 allows foran improved assessment of TMB. Whilst tNGS on fresh POSTER PRESENTATIONS S374 Journal of Hepatology 2020 vol. 73 | S123–S400