DETECT: Development of Technologies for Early HCC Detection

Abstract

Hepatocellular carcinoma (HCC) is a leading cause of morbidity and mortality in patients with cirrhosis and the third most common cause of cancer-related deaths worldwide.1Sung H. Ferlay J. Siegel R.L. et al.Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries.CA Cancer J Clin. 2021; 71: 209-249Crossref PubMed Scopus (26505) Google Scholar Though considered to be a highly fatal cancer, with a 5-year survival rate of less than 15%, curative treatments (surgical resection, liver transplantation, ablation) are available for patients diagnosed at an early stage. U.S. professional society guidelines recommend HCC surveillance in patients with cirrhosis of any etiology with biannual ultrasound ± serum alpha-fetoprotein (AFP),2Tzartzeva K. Obi J. Rich N.E. et al.Surveillance imaging and alpha fetoprotein for early detection of hepatocellular carcinoma in patients with cirrhosis: a meta-analysis.Gastroenterology. 2018; 154: 1706-1718.e1Abstract Full Text Full Text PDF PubMed Scopus (404) Google Scholar which is associated with survival benefits and receipt of curative treatment. However, ultrasound ± AFP can be associated with physical harms, that is, the need for multiple contrast-enhanced cross-sectional imaging studies and biopsies driven by false-positive tests or indeterminate lesions.3Atiq O. Tiro J. Yopp A.C. et al.An assessment of benefits and harms of hepatocellular carcinoma surveillance in patients with cirrhosis.Hepatology. 2017; 65: 1196-1205Crossref PubMed Scopus (135) Google Scholar Moreover, several new challenges in HCC surveillance are emerging. Contemporary Western cirrhosis cohorts consist of a higher proportion of patients with hepatitis C virus (HCV) after sustained virologic response and nonalcoholic fatty liver disease (NAFLD) with a lower annual risk of HCC (1%–3%), compared with older cohorts with viremic HCV (3%–8%). However, given the high prevalence of NAFLD in the U.S., the population at risk is growing.4Singal A.G. El-Serag H.B. Rational screening approaches for HCC in NAFLD patients.J Hepatol. 2022; 76: 195-201Abstract Full Text Full Text PDF PubMed Scopus (26) Google Scholar Given cost and capacity concerns, screening all patients with NAFLD is not feasible. Similar challenges exist for patients with HCV-related cirrhosis who have been cured. In addition, ultrasound has several limitations, including operator dependence, limited visualization in selected patients (particularly in the setting of abdominal adiposity and hepatic steatosis), and low sensitivity for early HCC (∼45%). Furthermore, given logistical barriers, imaging-based surveillance continues to be underutilized in clinical practice, with utilization rates of only 24%.5Wolf E. Rich N.E. Marrero J.A. et al.Use of hepatocellular carcinoma surveillance in patients with cirrhosis: a systematic review and meta-analysis.Hepatology. 2021; 73: 713-725Crossref PubMed Scopus (83) Google Scholar Therefore, the current “one-size-fits-all” model of ultrasound-based surveillance should be challenged and the discovery of new approaches and biomarkers are warranted. Biomarkers, as defined by the National Cancer Institute (NCI), are biological molecules found in blood, other body fluids, or tissues that are a sign of a normal or abnormal process, or of a condition or disease. In the present review, we discuss imaging biomarkers (ie, data collected by imaging techniques). There are increasing efforts to identify and validate novel biomarkers for risk stratification, early detection, diagnosis, prognostication, and evaluation of treatment response to improve outcomes for patients with cancer. For HCC early detection, an ideal biomarker should be simple, noninvasive, low cost, and widely available, and should have adequate sensitivity and specificity. The 5 phases of biomarker development extend from early biomarker discovery (phase I) and evaluation of biomarker performance (phases II and III) to prospective cohort and randomized studies (phases IV and V) in the clinical setting to evaluate benefits and harms.6Singal A.G. Hoshida Y. Pinato D.J. et al.International Liver Cancer Association (ILCA) white paper on biomarker development for hepatocellular carcinoma.Gastroenterology. 2021; 160: 2572-2584Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar Experimental design in biomarker studies from discovery to validation is key. The Early Detection Research Network (EDRN) was one of the early frameworks set forth by the NCI to address the 5 clinical trial phases for biomarker discovery for all cancer.7Pepe M.S. Etzioni R. Feng Z. et al.Phases of biomarker development for early detection of cancer.J Natl Cancer Inst. 2001; 93: 1054-1061Crossref PubMed Scopus (1241) Google Scholar An International Liver Cancer Association (ILCA) white paper adapted the EDRN framework to address HCC-specific considerations.6Singal A.G. Hoshida Y. Pinato D.J. et al.International Liver Cancer Association (ILCA) white paper on biomarker development for hepatocellular carcinoma.Gastroenterology. 2021; 160: 2572-2584Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar The goals of early phase I/II studies are discovery to determine how well the novel biomarker distinguishes HCC from non-HCC control specimens (ie, true positive rate and false positive rate), and development of clinical assays using specimens obtained noninvasively. Although a biomarker may appear to be promising in the exploratory phase, most fail in later phases owing to technical variability in assays. Phase II study design should include testing of the performance of the novel biomarker compared with the current standard of care (ie, ultrasound ± AFP).6Singal A.G. Hoshida Y. Pinato D.J. et al.International Liver Cancer Association (ILCA) white paper on biomarker development for hepatocellular carcinoma.Gastroenterology. 2021; 160: 2572-2584Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar AFP is the best studied and only biomarker to go through all 5 phases of validation. Though widely available and inexpensive, AFP alone has limited sensitivity for early-stage HCC and is prone to false positives in the setting of active viral hepatitis. More recent algorithms, such as GALAD (gender, age, AFP-L3, and des-gamma-carboxy prothrombin), have achieved higher sensitivity, exceeding 70%, in phase II/III studies and are discussed in depth elsewhere.8Singal A.G. Nabihah T. Mehta A. et al.GALAD demonstrates high sensitivity for HCC surveillance in a cohort of patients with cirrhosis.Hepatology. 2022; 75: 541-549Crossref PubMed Scopus (20) Google Scholar In this review, we discuss advances and challenges (Table 1) in novel biomarker development for HCC early detection, with a focus on those currently in very early discovery phases (I/II) of evaluation: cell-free DNA (cfDNA), circulating tumor cells (CTCs), high-throughput proteomic, metabolomic, and lipidomic platforms, radiomics approaches that utilize machine and deep learning (DL), and near-infrared fluorescence (NIRF) imaging.Table 1Summary of Advantages and Challenges of New Modalities for Hepatocellular Carcinoma (HCC) Early DetectionModalityAdvantagesChallengesOpportunities/future directionscfDNA•Fast processing time•Capture multiple biological features from tumor-derived DNA•Minimally invasive•Blood based (peripheral draw)•Low abundances, not detectable in very early cancers <1 cm in size•Short life/low stability•Biological noise in healthy control individuals•Improve assay sensitivity (eg, AS-PCR, COLD-PCR)•Need large cohorts of early HCC (≤2 cm) for validation•Combine with other markers to improve sensitivity and specificityCTCs•Diagnosis and monitoring disease progression•Fast processing time•Low concentrations•Proportional to tumor volume; more abundant in later-stage HCC•Heterogeneity in detection assays•Low specificity•Improve assay sensitivity•HCC-specific surface markers for detection•Need large cohorts of early HCC (≤2 cm) for validation•Combine with other markers to improve sensitivity and specificityEVs•Small volume needed•High specificity•Heterogeneity in assays•Need large cohorts of early HCC (≤2 cm) for validation•Assessment in other readily available body fluids•Combine with other markers to improve sensitivity and specificityProteomic, metabolomic, and lipidomic•Detect unique features of NAFLD-HCC•Heterogeneity in case and control specimens, assays, and calibration•Large sample sizes needed•Developing omics-based approaches to standardize methodologies across populations, controls, and data analysis•Combine with other markers to improve sensitivity and specificityConventional radiomics•Noninvasive•Use imaging data that is already available•Both surveillance and diagnosis•Need standardization•High variabilities in imaging protocols between centers•Require access to large sample size, high-quality training sets•Need to address applicability and generalizability to clinical practice•Need multi-center clinical studies and well annotated data sets to validate findings•Combine with liquid biomarkers to improve sensitivity and specificityDeep learning•Mitigates need for accurate segmentation, choice of user-defined featuresNIRF-probe imaging•Faster tumor uptake•High sensitivity due to deeper penetration depth in small tumors•Higher spatial resolution•High false-positive rate due to low specificity•Untargeted tissue distribution•Variability in serum stability•Highly HCC-specific probes•Serum stabilityAS-PCR, allele-specific polymerase chain reaction; cfDNA, cell-free DNA; COLD-PCR, co-amplification at lower denaturation temperature polymerase chain reaction; CTCs, circulating tumor cells; EVs, extracellular vesicles; NAFLD, nonalcoholic fatty liver disease; NIRF, near-infrared fluorescence. Open table in a new tab AS-PCR, allele-specific polymerase chain reaction; cfDNA, cell-free DNA; COLD-PCR, co-amplification at lower denaturation temperature polymerase chain reaction; CTCs, circulating tumor cells; EVs, extracellular vesicles; NAFLD, nonalcoholic fatty liver disease; NIRF, near-infrared fluorescence. Cell-free DNA in the blood, consisting in part of circulating tumor-derived DNA (ctDNA), are promising biomarkers for HCC early detection. The characteristic methylation patterns, mutations, and somatic copy number aberrations of ctDNA have been studied extensively for HCC early detection, with areas under the receiver operating characteristic curve (AUCs) ranging from 0.86 to 0.98 for early HCC.9Tran N.H. Kisiel J. Roberts L.R. Using cell-free DNA for HCC surveillance and prognosis.JHEP Rep. 2021; 3: 100304Abstract Full Text Full Text PDF PubMed Scopus (19) Google Scholar In an early-phase study, compared with patients with cirrhosis, the plasma methylated SEPT9 had a diagnostic accuracy AUC of 0.86 in patients with Barcelona Clinic Liver Cancer (BCLC) stage 0–A HCC.10Oussalah A. Rischer S. Bensenane M. et al.Plasma mSEPT9: a novel circulating cell-free DNA–based epigenetic biomarker to diagnose hepatocellular carcinoma.EBioMedicine. 2018; 30: 138-147Abstract Full Text Full Text PDF PubMed Scopus (82) Google Scholar Genome-wide 5-hydroxymethylcytosines in cfDNA samples outperformed AFP alone for early HCC detection, with AUCs of 0.92 to 0.88 in training and validation sets. It also demonstrated a high capacity for distinguishing patients with small HCC (defined as <2.0 cm) from non-HCC or patients with hepatitis B virus (HBV) infection/cirrhosis.11Cai J. Chen L. Zhang Z. et al.Genome-wide mapping of 5-hydroxymethylcytosines in circulating cell-free DNA as a noninvasive approach for early detection of hepatocellular carcinoma.Gut. 2019; 68: 2195-2205Crossref PubMed Scopus (130) Google Scholar The technical challenges of using cfDNA for HCC early detection need to be considered (Table 1). For example, different storage and extraction procedures of cfDNA from blood samples (ie, plasma vs serum) introduce significant variation in cfDNA quality. Therefore, standardizing the blood sample processing workflow will significantly facilitate the downstream interpretation of the results, especially for samples from different sources, and help to distinguish the cancer-associated mutations from the noise found in healthy control specimens. In addition, a precise quantitative analysis of cfDNA level for diagnostic accuracy of HCC has not been established. Extracellular vesicles (EVs) are vesicles found in the extracellular space that act as cargos for proteins, nucleic acids, apoptotic bodies, and other cellular parts released by normal and cancer cells.12Kosaka N. Kogure A. Yamamoto T. et al.Exploiting the message from cancer: the diagnostic value of extracellular vesicles for clinical applications.Exp Mol Med. 2019; 51: 1-9Crossref PubMed Scopus (73) Google Scholar Owing to the stability of EVs, their ability to be detected in many body fluids, and informing the cell of origin, their role as biomarkers has been investigated, including in HCC early detection.12Kosaka N. Kogure A. Yamamoto T. et al.Exploiting the message from cancer: the diagnostic value of extracellular vesicles for clinical applications.Exp Mol Med. 2019; 51: 1-9Crossref PubMed Scopus (73) Google Scholar For example, the sensitivity, specificity, and AUC for exosome-derived lactate dehydrogenase C4 mRNA were 88.2%, 93.3%, and 0.945, respectively, in distinguishing early-stage HCC patients (TNM stage I/II) from healthy control specimens.13Cui Z. Li Y. Gao Y. et al.Cancer-testis antigen lactate dehydrogenase C4 in hepatocellular carcinoma: a promising biomarker for early diagnosis, efficacy evaluation and prognosis prediction.Aging (Albany NY). 2020; 12: 19455-19467Crossref PubMed Scopus (17) Google Scholar EV Click Chips, which combine immunoaffinity with microfluidic devices, were recently developed and optimized to purify HCC-specific EVs. With them, a 10-marker gene HCC-specific panel showed great potential in distinguishing patients with early HCC (defined as BCLC stage 0–A), compared with patients with cirrhosis, with 94.4% sensitivity, 88.5% specificity, and AUC 0.93.14Sun N. Lee Y.T. Zhang R.Y. et al.Purification of HCC-specific extracellular vesicles on nanosubstrates for early HCC detection by digital scoring.Nat Commun. 2020; 11: 4489Crossref PubMed Scopus (85) Google Scholar Validation cohorts are needed to compare ultrasound ± AFP vs EV-based methods for early-stage tumors <2 cm in size. In addition, given the technical variability in EV purification/isolation methods, standardization of protocols will also be critical in selecting HCC-specific derived cargos to be tested clinically. Circulating tumor cells (CTCs), like cfDNA, can be detected noninvasively via a simple peripheral blood draw.15Ahn J.C. Teng P.C. Chen P.J. et al.Detection of circulating tumor cells and their implications as a biomarker for diagnosis, prognostication, and therapeutic monitoring in hepatocellular carcinoma.Hepatology. 2021; 73: 422-436Crossref PubMed Scopus (108) Google Scholar CTCs are present in very low concentrations in the blood (1–10 CTCs per mL), and the number of CTCs is positively correlated with the disease stage. Although CTCs have shown promise as a prognostic biomarker in patients with HCC to predict recurrence and treatment response, emerging evidence also indicates that CTCs can be used for early detection with evolving technologies for CTCs isolation and detection. For example, immunoaffinity-based enrichment techniques and microfluidics in the CTC-iChip may allow for CTC detection rates 100-fold higher than the FDA-approved CellSearch CTC kit.16Wan S. Kim T.H. Smith K.J. et al.New Labyrinth microfluidic device detects circulating tumor cells expressing cancer stem cell marker and circulating tumor microemboli in hepatocellular carcinoma.Sci Rep. 2019; 9: 18575Crossref PubMed Scopus (31) Google Scholar A recent study showed that the Labyrinth microfluidic device, combining HCC biomarkers (glypican-3, glutamine synthase, and Hep Par-1) with the cancer stem cell marker CD44, detected early-stage HCC (TNM 0–I) with a 75% positive rate and more advanced stages (TNM II–IV) at 95%.16Wan S. Kim T.H. Smith K.J. et al.New Labyrinth microfluidic device detects circulating tumor cells expressing cancer stem cell marker and circulating tumor microemboli in hepatocellular carcinoma.Sci Rep. 2019; 9: 18575Crossref PubMed Scopus (31) Google Scholar Similarly, CanPatrol enrichment using the RNA-ISH assay successfully detected CTCs in early-stage and late-stage HCC with true positive rates of 83.6% and 96.5%, respectively, while no CTCs were detected in healthy donors17Qi L.N. Xiang B.D. Wu F.X. et al.Circulating tumor cells undergoing EMT provide a metric for diagnosis and prognosis of patients with hepatocellular carcinoma.Cancer Res. 2018; 78: 4731-4744Crossref PubMed Scopus (147) Google Scholar; however, it has failed to differentiate HCC from other cancer types and is unable to detect CTCs that do not express epithelial cellular adhesion molecule (EpCAM). In addition, there are technical variabilities in the detection and isolation of CTCs, a complex process that typically begins with physical isolation (size, velocity, density) followed by isolation based on biological aspects of CTCs, that is, surface markers of epithelial or mesenchymal origin. Therefore, current CTC assays and data are heterogeneous, and standardization of CTC assays is needed. Despite a favorable performance with diagnostic accuracy for early-stage HCC, these studies using blood-derived cfDNA, EVs, or CTCs are limited by the number of early-stage HCC patients (BCLC 0–A with small HCC <2 cm) and the heterogeneity in tumor sizes for BCLC A-stages. Prospective validation in large cohorts of BCLC 0–A HCC patients will be needed to improve diagnostic accuracy of cfDNAs, EVs, and CTCs in HCC <2 cm and to improve their clinical application toward personalized HCC surveillance. Although labs-on-chips, microfluidic chips, and proximity extension assays allow simultaneous detection of a few to hundreds of cancer-related protein biomarkers, current liquid biomarker studies may be greatly improved by high-throughput proteomics platforms.18Liu R.X. Thiessen-Philbrook H.R. Vasan R.S. et al.Comparison of proteomic methods in evaluating biomarker-AKI associations in cardiac surgery patients.Transl Res. 2021; Abstract Full Text Full Text PDF Scopus (11) Google Scholar Significant alterations in serum levels of metabolites (eg, phosphatidylcholine, sphingosine, lysophosphatidylcholine, serine, glycine, and aspartate) have shown the ability to discriminate between HCC and cirrhosis in an HBV population in the discovery, test, and validation sets.19Luo P. Yin P. Hua R. et al.A large-scale, multicenter serum metabolite biomarker identification study for the early detection of hepatocellular carcinoma.Hepatology. 2018; 67: 662-675Crossref PubMed Scopus (201) Google Scholar A serum metabolite biomarker panel consisting of phenylalanyl-tryptophan and glycocholate outperformed AFP in detecting HCC ≤3 cm (AUC 0.87 vs 0.68) vs those with cirrhosis. Moreover, the panel accurately diagnosed HCC in 90.5% of patients with a false-negative AFP and 75.0% of patients with small HCCs ≤ 3 cm (AFP <20 ng/mL).19Luo P. Yin P. Hua R. et al.A large-scale, multicenter serum metabolite biomarker identification study for the early detection of hepatocellular carcinoma.Hepatology. 2018; 67: 662-675Crossref PubMed Scopus (201) Google Scholar Although metabolomic signature patterns have been identified for early detection for HBV- and HCV-associated HCC,20Zhou L. Ding L. Yin P. et al.Serum metabolic profiling study of hepatocellular carcinoma infected with hepatitis B or hepatitis C virus by using liquid chromatography–mass spectrometry.J Proteome Res. 2012; 11: 5433-5442Crossref PubMed Scopus (61) Google Scholar data in NAFLD patients are conflicting. Serum and intrahepatic lipid compositions do not always correlate in NAFLD/non-alcoholic steatohepatitis (NASH); however, studies in NAFLD-related HCC show distinct intrahepatic and serum lipid profiles, supporting the idea that intrahepatic lipid content may not reflect what is detected in the serum of HCC patients. Yet serum lipid signatures have shown promise in case-control studies for diagnosing HCC and prognostication,21Lu Y. Chen J. Huang C. et al.Comparison of hepatic and serum lipid signatures in hepatocellular carcinoma patients leads to the discovery of diagnostic and prognostic biomarkers.Oncotarget. 2018; 9: 5032-5043Crossref PubMed Scopus (27) Google Scholar with few studies addressing their role in early detection. In a case-control study of NASH-HCC murine models, followed by validation in human NASH-HCC specimens, HCC cases not only had a distinct serum lipid signature compared with the NASH control specimens, but the HCC lipid profile was also associated with tumor burden.22Muir K. Hazim A. He Y. et al.Proteomic and lipidomic signatures of lipid metabolism in NASH-associated hepatocellular carcinoma.Cancer Res. 2013; 73: 4722-4731Crossref PubMed Scopus (112) Google Scholar In a prospective cohort study, n-3 and n-6 polyunsaturated fatty acids were identified in patients with cirrhosis who developed HCC.23Khan I.M. Gjuka D. Jiao J. et al.A novel biomarker panel for the early detection and risk assessment of hepatocellular carcinoma in patients with cirrhosis.Cancer Prev Res (Phila). 2021; 14: 667-674Crossref PubMed Scopus (3) Google Scholar Whether such lipid profiles are etiology specific and affected by medications such as statins will need to be evaluated. Imaging plays a central role in the surveillance, diagnosis, and treatment of HCC. HCC is unique among cancers in that the diagnosis can be made radiographically in a majority of cases, without the need for histologic confirmation, given the liver’s dual blood supply. However, interpretation and analysis by radiologists can also be subjective with poor inter-rater reliability. Recent advances in computer science have enabled the clinical application of computer-assisted analysis in imaging, which is more objective, replicable, and comprehensive. High-dimensional digital imaging biomarkers were extracted from cross-sectional imaging studies, which expand big-data “omics” to the imaging domain. These quantitative radiologic data can then be leveraged to characterize patient heterogeneity and build precision medicine algorithms for diagnostic and predictive tasks. Radiomics needs manual segmentation by radiologists, which limits its use as a screening tool.24Harding-Theobald E. Louissaint J. Maraj B. et al.Systematic review: radiomics for the diagnosis and prognosis of hepatocellular carcinoma.Aliment Pharmacol Ther. 2021; 54: 890-901Crossref PubMed Scopus (20) Google Scholar In comparison, DL, especially that based on convolutional neural networks, has been applied to medical image analysis, transforming raw images to more abstract and composite representations for prediction and classification. Therefore, DL does not rely on hand-engineered features as radiomics does. Several studies have shown that these models can improve on early radiologic characterization and diagnostic accuracy of HCC through computer-aided diagnoses. Although imaging data have mostly focused on HCC diagnosis, especially in more advanced stages or disease recurrence, including the current Liver Imaging Reporting and Data System,25Chernyak V. Fowler K.J. Kamaya A. et al.Liver Imaging Reporting and Data System (LI-RADS) version 2018: imaging of hepatocellular carcinoma in at-risk patients.Radiology. 2018; 289: 816-830Crossref PubMed Scopus (413) Google Scholar efforts are still ongoing to develop novel algorithms for HCC early detection to overcome some limitations of current surveillance.26Zhen S.H. Cheng M. Tao Y.B. et al.Deep learning for accurate diagnosis of liver tumor based on magnetic resonance imaging and clinical data.Front Oncol. 2020; 10: 680Crossref PubMed Scopus (68) Google Scholar Combining with recurrent neural network models, DL can integrate medical images with existing electronic health records (EHRs) and improve the prediction of clinical events. It has the advantage of capturing temporal dynamic information and predicting the risk for developing HCC, which fluctuates over time.27Ioannou G.N. Tang W. Beste L.A. et al.Assessment of a deep learning model to predict hepatocellular carcinoma in patients with hepatitis C cirrhosis.JAMA Netw Open. 2020; 3e2015626Crossref Scopus (37) Google Scholar Ioannou et al demonstrated that DL models using longitudinal Veterans Health Administration EHR data outperformed logistic regression models in predicting the risk for HCC in patients with viral hepatitis and alcohol-associated liver disease.27Ioannou G.N. Tang W. Beste L.A. et al.Assessment of a deep learning model to predict hepatocellular carcinoma in patients with hepatitis C cirrhosis.JAMA Netw Open. 2020; 3e2015626Crossref Scopus (37) Google Scholar Despite these advantages, one major limitation is that large datasets with thousands of cases are needed for training the models to achieve high performance. The complexity of deep layers limits the interpretability of how neural networks make decisions for predictions. This “blackboxness” can be troubling for the field of health care and needs further investigation for clinical implementation.28Gillies R.J. Schabath M.B. Radiomics improves cancer screening and early detection.Cancer Epidemiol Biomarkers Prev. 2020; 29: 2556-2567Crossref PubMed Scopus (41) Google Scholar Furthermore, the application of DL toward radiomics, which analyzes phenotypic and quantitative features from radiographic images, is an emerging technology for early detection and prognostication in several other cancers. In HCC, various studies have demonstrated the role of radiomics in diagnosis and prognostication of treatment response and recurrence. However, the variability in image quality and the lack of cohorts with early-stage HCC lesions with longitudinal radiologic imaging has limited its applicability for early detection. Since the application of near-infrared indocyanine green fluorescence in identifying tumors in liver cancer surgery, various preclinical NIRF probes (eg, peptide-based ligands), which bind specifically to surface proteins highly expressed in early-stage HCC (eg, Sp17, GPC3, integrin αvβ3 receptor, EGFR), have been developed for in vivo imaging of HCC.29Zhang W. Hu Z. Tian J. et al.A narrative review of near-infrared fluorescence imaging in hepatectomy for hepatocellular carcinoma.Ann Transl Med. 2021; 9: 171Crossref PubMed Google Scholar The advantages of NIRF probes include faster tumor uptake, deeper penetration depth owing to smaller size/molecular weight, higher spatial resolution, and sensitivity. However, a high false-positive rate was reported due to low specificity. Currently, more HCC-specific peptides labeled with MRI/CT/PET contrast agents are being developed to identify tumors.29Zhang W. Hu Z. Tian J. et al.A narrative review of near-infrared fluorescence imaging in hepatectomy for hepatocellular carcinoma.Ann Transl Med. 2021; 9: 171Crossref PubMed Google Scholar For HCC early detection, advances need to be made to increase the specificity of NIRF probes for HCC, serum stability and minimize untargeted tissue distribution. The aforementioned biomarkers are all currently being evaluated by the NCI-funded Translational Liver Consortium (TLC).30Singal A.G. Lok A.S. Feng Z. et al.Conceptual model for the hepatocellular carcinoma screening continuum: current status and research agenda.Clin Gastroenterol Hepatol. 2020; Abstract Full Text Full Text PDF Scopus (26) Google Scholar Key design principles that investigators should consider include defining the target population and intended clinical use of the biomarker, ensuring specimen collection before outcome ascertainment, specifying procedures for specimen collection, processing, storage, and retrieval, blinding of specimens to outcome status, and evaluation of the performance of the biomarkers based on the intended clinical use. Principles of biomarkers discovery pipelines should also be followed as delineated by the NCI EDRN PRoBE (prospective specimen collection, retrospective blinded evaluation) design and the ILCA white paper on biomarker development for HCC,6Singal A.G. Hoshida Y. Pinato D.J. et al.International Liver Cancer Association (ILCA) white paper on biomarker development for hepatocellular carcinoma.Gastroenterology. 2021; 160: 2572-2584Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar with eventual validation in prospective cohorts and randomized studies.6Singal A.G. Hoshida Y. Pinato D.J. et al.International Liver Cancer Association (ILCA) white paper on biomarker development for hepatocellular carcinoma.Gastroenterology. 2021; 160: 2572-2584Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar,7Pepe M.S. Etzioni R. Feng Z. et al.Phases of biomarker development for early detection of cancer.J Natl Cancer Inst. 2001; 93: 1054-1061Crossref PubMed Scopus (1241) Google Scholar In addition, including more than 1 independent validation cohort and ensuring sufficiently powered studies, especially for DL/machine learning, will also increase the likelihood that a biomarker (or algorithm) can move from the early phases of biomarker development into clinical use. To date, most biomarkers remain in the early discovery phases (phase I/II). Currently, among the blood-based biomarkers in earlier stages of development, ctDNA has the highest potential to be used in the clinical setting for HCC early detection in the near future, given the stability of DNA in the blood, the relative ease of purification and analytic methods, and the potential of ctDNA to identify the tissue of origin.31Li S. Noor Z.S. Zeng W. et al.Sensitive detection of tumor mutations from blood and its application to immunotherapy prognosis.Nat Commun. 2021; 12: 4172Crossref PubMed Scopus (9) Google Scholar Whereas the progression of EVs and CTCs to the next phase of clinical trial design is hindered by their challenging methodologies (eg, lack of standardized and reproducible isolation method and poor specificity). As the field of early HCC detection evolves and assays are refined, one potential application of these novel biomarkers is developing a multi-tiered approach in HCC surveillance. For example, at-risk patients would be identified with the use of clinically embedded EHR DL/machine learning algorithms that use common clinical parameters and radiomics data. Patients who are deemed “high risk” would subsequently be evaluated with validated serum biomarkers, prompting further cross-sectional imaging (Figure 1). Because HCC growth patterns are heterogeneous,32Rich N.E. John B.V. Parikh N.D. et al.Hepatocellular carcinoma demonstrates heterogeneous growth patterns in a multi-center cohort of patients with cirrhosis.Hepatology. 2020; Crossref Scopus (68) Google Scholar,33Benhammou J.N. Lin J. Aby E.S. et al.Nonalcoholic fatty liver disease-related hepatocellular carcinoma growth rates and their clinical outcomes.Hepatoma Res. 2021; 7PubMed Google Scholar optimal screening intervals (eg, slow- vs fast-growing tumors based on biomarker thresholds) will need to be addressed, as well as management of patients with positive serum biomarkers in the absence of radiographic HCC findings. Synergism between the clinical and scientific communities is critical to the continued advancement of HCC early detection biomarker discovery from bench to bedside. We thank Drs Anna Lok, Amit Singal, and Jo Ann S Rinaudo for their valuable feedback and input during the writing of the commentary. We also acknowledge the National Institutes of Health/National Cancert Institute Translational Liver Consortium for supporting the Early-Career Working Group (U01CA230669, U01CA230690, U01CA230694, U01CA230705, U01CA230997), including administrative support from Lesley Wilkerson and Royceann Malnik. The unlisted authors from the National Cancer Institute’s Translational Liver Consortium Early-Career Working Group include: Wenyuan Li (Department of Pathology and Laboratory Medicine, University of California, Los Angeles, California), Xiaoli Wu (Division of Gastroenterology, Department of Internal Medicine, University of Michigan, Ann Arbor, Michigan), Shuo Feng (Division of Gastroenterology, Department of Internal Medicine, University of Michigan, Ann Arbor, Michigan), Naoto Fujiwara (University of Texas Southwestern Medical Center, Dallas, Texas); Xiaoqing Meng (Division of Gastroenterology, Department of Internal Medicine, University of Michigan, Ann Arbor, Michigan), Shijia Zhu (University of Texas Southwestern Medical Center, Dallas, Texas); Sobia Zaidi (Department of Biomedical Sciences, Ohio University, Athens, Ohio). Opinions expressed by the authors are their own; this material should not be interpreted as representing the official viewpoint of the U.S. Department of Health and Human Services, the National Institutes of Health, or the National Cancer Institute.