Gastroenterology
Volume 137, Issue 1 , Pages 26-29, July 2009

α-Fetoprotein for Hepatocellular Carcinoma Diagnosis: The Demise of a Brilliant Star

  • Alejandro Forner
    • ,
  • María Reig
    • ,
  • Jordi Bruix

      Affiliations

    • Corresponding Author InformationReprint requests Address requests for reprints to: Jordi Bruix, MD, BCLC Group, Liver Unit, Hospital Clinic, IDIPAPS, CIBEREHD, University of Barcelona, c/ Villarroel 170, E-08036, Barcelona, Spain. fax: +34 93 227 57 92

BCLC Group, Liver Unit, Hospital Clinic, Institut d'Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Centro de Investigación Biomédica en Red de Enfermedades Hepáticas y Digestivas (CIBEREHD), University of Barcelona, Barcelona, Spain

published online 01 June 2009.

Article Outline

 

See “α-Fetoprotein, Des-γ Carboxyprothrombin, and lectin-bound α-fetoprotein in early hepatocellular carcinoma,” by Marrero JA, Feng Z, Wang Y, et al, on page 110.

Until recently, hepatocellular carcinoma (HCC) was considered a geographically restricted cancer that affected specific regions of the world where a combination of viral infections, aflatoxin intake and alcohol consumption are highly prevalent. This has changed: HCC has a higher prevalence in South East Asia and sub-Saharan Africa, but its incidence has increased significantly in Europe and America, where the main risk factors are hepatitis C virus infection, alcoholism, and the metabolic syndrome.1, 2 Indeed, several studies have shown that HCC is now the leading cause of death in cirrhotic patients; hence, it has become a major problem in this at-risk population.3

It has been long maintained that HCC is large, multifocal, and highly symptomatic, so that diagnosis at an early stage was not feasible and treatment difficult. Even operative resection for solitary HCC in compensated cirrhosis was challenged because improvements in survival were sometimes felt to be marginal. The established observation was that the so-called minute HCC was a specific entity described in Asia, but not existing in the West. However, now early HCC may be detected with proper imaging, it may be diagnosed at a fully asymptomatic stage even in the absence of relevant tumor burden, and patients may benefit from several treatment options with positive impact in survival if properly applied. These include operative resection, liver transplantation, ablation, chemoembolization, and sorafenib administration.4

The demonstration of the benefits of a surveillance policy should be derived from a randomized controlled investigation comparing surveillance versus no surveillance. Its cost effectiveness will be mostly related to the sensitivity and specificity of the surveillance tools, as well as the efficacy of treatment. Does such a randomized study exist for HCC? And do the surveillance tools have sufficient accuracy? There is a single, randomized control trial assessing the benefits of surveillance. It included Chinese patients with hepatitis B virus (HBV) infection and, despite several limitations (heterogeneity of the population, lack of adherence, lack of uniform treatment approach), survival was significantly improved in the surveillance cohort.5 The remaining data supporting the benefits of surveillance for HCC in patients with cirrhosis come from cohort investigations and similar suboptimal assessments.6, 7, 8 However, it is widely acknowledged that a proper randomized control trial is not feasible in settings where health care is of adequate quality. The tools used for surveillance are ultrasound (US) examination and tumor marker determination; because they are part of the conventional clinical evaluation of patients with chronic liver disease, even if not aimed at screening for HCC, any trial would be spoiled by unscheduled examinations by the same tools and distort the study itself. In addition, because the sole way to apply effective therapy for HCC is to diagnose it at an early stage, patients at risk of HCC (namely patients with cirrhosis) would refuse to enter the trial if properly informed. Hence, such a trial is no longer a feasible option. Unfortunately, the absence of strong data to support screening unequivocally for HCC in patients at risk fuels the debate as to what extent it should be accepted as standard of care, and if it should be even recommended. Although, from an intellectual point of view, the debate is to be accepted, the current position and state-of-the-art thinking panels of experts and scientific guidelines are to recommend screening for HCC, being fully aware of existing limitations.9, 10 These include not only the already mentioned absence of data produced within proper randomized controlled trials, but also <100% detection efficacy of the tools to be employed for surveillance.

Surveillance tools include tumor marker assessment and US examination. The first is assessed in the paper by Marrero et al11 and is the primer for this editorial. However, at the same time they argue against the usefulness of US for surveillance and claim that it is not sensitive or specific, thus making it necessary to develop new markers and refine the application of those that are currently available. To frame the debate, it is worth mentioning that US is highly operator dependent and hence, if not applied by skilled individuals, its performance will never be competitive. Thus, to dismiss its value because of poor expertise may not be acceptable; the same would apply to other imaging techniques, not to mention treatment application such as resection, transplantation, or ablation, where US is also required for proper performance. Obviously, interpretation of suspicious images is challenging and this is especially relevant in morbid obesity or in advanced cirrhosis with severely distorted liver architecture. However, surveillance of end-stage cirrhosis with an atrophic nodular liver is not the setting where surveillance offers benefits. All practice guidelines state that only cirrhotic patients who would be treated if diagnosed with HCC should undergo surveillance; this excludes patients with end-stage cirrhosis who should be considered for transplantation, even without HCC development. If not considered already for transplantation because of any comorbidity, early HCC diagnosis will not be followed by therapy and no benefit will be derived from surveillance. Proper training and skills in countries such as Japan have shown that US is efficacious and a majority of HCC in that country are detected at an early stage when tumor eradication is feasible. Data from surveillance approaches in Europe show a sensitivity between 65% and 80% and a specificity >90%.6 In addition, upon detection of a suspicious nodule, the strategy for diagnosis confirmation has been established and validated even in small tumors.10, 12 Thus, the performance of US has room for improvement, but its value cannot be neglected, nor should its limitations be used to necessarily favor tumor markers.

Tumor markers have been used widely and evaluated as a potential diagnostic tool, while their usefulness as for screening tool is less characterized. The most widely established tumor marker is alpha-fetoprotein (AFP). Initially described by Abelev et al in 1963,13 it was the sole tool for raising suspicion of HCC when imaging techniques such as US were not available. Accordingly, in areas with high prevalence of HCC such as China, early detection plans were based on repeated AFP determinations.14 As expected, all HCC detected during screening were AFP positive. Those without an AFP increase would only be evaluated when symptomatic and this is why in some reports the proportion of HCC with increased AFP is higher at an early stage than a more advanced phase. The availability of a serologic test that could detect or raise the suspicion of HCC primed its fast implementation as a parameter to include in the biochemical evaluation of patients with liver disease. The presence of an increased concentration would trigger further examination by hepatic scintigraphy, laparoscopy, or angiography and in most cases an HCC could be diagnosed. At the same time of this sensitivity validation, it was soon recognized that there were increases in false-positive results. In the absence of malignancy an AFP increase could be related to flares in HBV replication15 and also of HCV.16 However, AFP could also increase in other malignancies. Some gastrointestinal cancers may show increased AFP levels17; cholangiocarcinoma may also present with elevated values.18 This is important to remark as cholangiocarcinoma occurs also in the setting of cirrhosis and has a completely different prognosis and treatment. As a whole, the specificity of an abnormal AFP concentration will never be 100% and when the development of imaging techniques have allowed the detection of HCC at an early stage, the usefulness of AFP has progressively decreased. In a large case-control study by Trevisani et al,19 the limited value of AFP in the clinical setting was fully exposed. Even if using the best cutoff according to receiver operating characteristics curve (20 ng/ml) the sensitivity did not exceed 60%, this meaning a 40% of false negatives, and a mere 90% specificity was obtained.19

The study by Marrero et al involves a large, American cohort and includes in the assessment 2 other tumor markers: des-gamma-carboxy prothrombin (DCP) and lens culinaris agglutin-reactive AFP (L3-AFP). A total of 836 patients were enrolled, 417 (50%) were control cirrhotics without HCC and 419 (50%) were HCC cases, of which 208 (49.6%) had early stage HCC (n = 77 very early, n = 131 early) according to the Barcelona Clinic Liver Cancer (BCLC) staging system. The data show that AFP has the best area under the receiver operating characteristic curve (0.80; 95% confidence interval, 0.77–0.84) for early stage HCC and that the optimal AFP cutoff value would have to be set at 10.9 ng/ml to provide a sensitivity of 66%. The combination of AFP and DCP improved the sensitivity to 70% for early stage HCC and to 78% in those with viral etiology and early stage HCC.11 The authors confirm that AFP is the best tumor marker for HCC diagnosis, and its performance improves when it is associated with DCP, particularly in those with viral cirrhosis. At the same time, they show that L3-AFP is not useful for early HCC diagnosis and DCP has the same diagnostic capacity as AFP.

The authors are to be commended for assembling such a large cohort to define the usefulness of tumor markers in their setting. The cohort includes all relevant stages of the disease at diagnosis, and proper mix of etiologies and patients profile so that several subgroup assessments are feasible, as well as a multivariate approach. The analysis using receiver operating characteristics curve permits the proposal of the best cutoff that for AFP is within the current normal concentration values. Finally, the assessment of not only AFP, but also L3-AFP and DCP, has added interest to this study.

The main conclusion is that AFP is the best tumor marker for early HCC diagnosis. DCP showed worse performance than AFP for early detection despite the promising results published by the same authors.20 Thus, its use as screening tool should be discouraged. Does this finding mean that AFP is a useful tool for surveillance either alone or in conjunction with the other markers? Several aspects suggest that the answer is no. The study by Marrero et al is an assessment of AFP for diagnosis and not for surveillance. The marker is assessed in patients with established HCC diagnosis and as such is a retrospective, case-control study (a Phase II study according to the National Cancer Institute's Early Detection Research Network). Hence, the final answer should come from prospective follow-up of a cohort of patients with cirrhosis to determine whether the biomarker can detect HCC cases before clinical diagnosis. If the sensitivity for diagnosis is clearly suboptimal and there is a correlation between AFP concentration and tumor stage, it has to be predicted that the sensitivity for surveillance will be even worse and, therefore, no major hope has to be placed in such an effort. Furthermore, the diagnostic accuracy of any test is related to the frequency of the underlying disease in the population being studied. Therefore, the diagnostic accuracy obtained in a case-control study, where there is a preestablished prevalence of 50%, cannot be reproduced when the test is applied for screening purposes in a population with significantly less disease frequency. In addition, a longitudinal design minimizes the presence of biases, mainly inclusion and verification biases. The authors state that the verification bias has been controlled because they did not use AFP as a diagnostic tool for defining case. However, they cannot ensure that AFP has not been used as screening, so in a significant proportion of individuals in the case group, HCC diagnosis may have been triggered by the finding of elevated AFP during surveillance, thus repeating the bias of the AFP-based surveillance studies. This could also justify why the performance of AFP is indeed better in early HCC compared with advanced HCC, when a higher tumor markers values should be expected because of the correlation between tumor stage and AFP as described by Marrero in previous studies.20 Probably, most of the non-early HCC cases may come from unscreened population or from those patients with repeatedly negative AFP values, limiting the strength of the current data.

Another key point is that the control group has not been matched with cases according to gender, age, ethnicity, etiology of liver disease, and degree of liver impairment. This flaw translates in a different distribution of these variables between cases and controls, as reflected in the Table 1 of Marrero's study.11 The authors have discarded the presence of cofounders by logistic regression analysis, but a better strategy would have been to recruit a carefully matched control group. In addition, other potential confounders have not been included in the analysis. The levels of alanine aminotransferase, which may act as a surrogate marker of inflammation and activity, have been repeatedly correlated with the AFP values.15, 19 Other missing parameters are viral status and previous or concomitant antiviral treatment. Elevation of AFP levels has been associated with HBV flares,15 and a recent, randomized, controlled trial has shown that the levels of AFP decreased after continuous, low-dose pegylated interferon treatment in patients with chronic hepatitis C infection.16 The heterogeneous distribution of these variables between the case and control groups may confound and significantly modify the results and conclusions. In addition to all these comments, the most relevant drawback of AFP is the 80% specificity of an abnormal concentration as per the proposed cutoff. This would mean that 20% of the cirrhotics undergoing follow-up should be examined by imaging techniques without a final HCC diagnosis. The cost of such failures would make any surveillance plan impossible to sustain.

In summary, the data from Marrero et al further establish that AFP has lost the competition as a screening tool for HCC and, hence, the assumption of AFP as a clinically meaningful instrument for screening and diagnosis is a concept that should be withdrawn from clinical decision making. Its value is limited in terms of sensitivity and specificity; imaging techniques in proper hands are more accurate. Hence, no more efforts have to be placed in trying to resuscitate a tool that received a deserved obituary years ago.21 Proper training in US and active investigation in the field of new technologies such as proteomics and glycomics, should be seen as the optimal strategy to finally identify the best surveillance instruments.

Back to Article Outline

References 

  1. El-Serag HB, Mason AC. Rising incidence of hepatocellular carcinoma in the United States. N Engl J Med. 1999;340:745–750
  2. Parkin DM, Bray F, Ferlay J, et al. Global cancer statistics, 2002. CA Cancer J Clin. 2005;55:74–108
  3. Sangiovanni A, Prati GM, Fasani P, et al. The natural history of compensated cirrhosis due to hepatitis C virus: a 17-year cohort study of 214 patients. Hepatology. 2006;43:1303–1310
  4. Bruix J, Llovet JM. Major achievements in hepatocellular carcinoma. Lancet. 2009;373:614–616
  5. Zhang B, Yang B. Combined alpha fetoprotein testing and ultrasonography as a screening test for primary liver cancer. J Med Screen. 1999;6:108–110
  6. Bolondi L, Sofia S, Siringo S, et al. Surveillance programme of cirrhotic patients for early diagnosis and treatment of hepatocellular carcinoma: a cost effectiveness analysis. Gut. 2001;48:251–259
  7. Arguedas MR, Chen VK, Eloubeidi MA, et al. Screening for hepatocellular carcinoma in patients with hepatitis C cirrhosis: a cost-utility analysis. Am J Gastroenterol. 2003;98:679–690
  8. Sarasin FP, Giostra E, Hadengue A. Cost-effectiveness of screening for detection of small hepatocellular carcinoma in western patients with Child-Pugh class A cirrhosis. Am J Med. 1996;101:422–434
  9. Bruix J, Sherman M, Llovet JM, et al. Clinical management of hepatocellular carcinoma. Conclusions of the Barcelona-2000 EASL conference. European Association for the Study of the Liver. J Hepatol. 2001;35:421–430
  10. Bruix J, Sherman M. Management of hepatocellular carcinoma. Hepatology. 2005;42:1208–1236
  11. Marrero JA, Feng Z, Wang Y, et al. α-Fetoprotein, Des-γ Carboxyprothrombin, and lectin-bound α-fetoprotein in early hepatocellular carcinoma. Gastroenterology. 2009;137:110–118
  12. Forner A, Vilana R, Ayuso C, et al. Diagnosis of hepatic nodules 20 mm or smaller in cirrhosis: Prospective validation of the noninvasive diagnostic criteria for hepatocellular carcinoma. Hepatology. 2008;47:97–104
  13. Abelev GI, Perova SD, Khramkova NI, et al. Production of embryonal alpha-globulin by transplantable mouse hepatoma. Transplantation. 1963;1:174–180
  14. Chen JG, Parkin DM, Chen QG, et al. Screening for liver cancer: results of a randomised controlled trial in Qidong, China. J Med Screen. 2003;10:204–209
  15. Lok AS, Lai CL. alpha-Fetoprotein monitoring in Chinese patients with chronic hepatitis B virus infection: role in the early detection of hepatocellular carcinoma. Hepatology. 1989;9:110–115
  16. Di Bisceglie AM, Sterling RK, Chung RT, et al. Serum alpha-fetoprotein levels in patients with advanced hepatitis C: results from the HALT-C Trial. J Hepatol. 2005;43:434–441
  17. Adachi Y, Tsuchihashi J, Shiraishi N, et al. AFP-producing gastric carcinoma: multivariate analysis of prognostic factors in 270 patients. Oncology. 2003;65:95–101
  18. Ishikawa K, Sasaki A, Haraguchi N, et al. A case of an alpha-fetoprotein-producing intrahepatic cholangiocarcinoma suggests probable cancer stem cell origin. Oncologist. 2007;12:320–324
  19. Trevisani F, D'Intino PE, Morselli-Labate AM, et al. Serum alpha-fetoprotein for diagnosis of hepatocellular carcinoma in patients with chronic liver disease: influence of HBsAg and anti-HCV status. J Hepatol. 2001;34:570–575
  20. Marrero JA, Su GL, Wei W, et al. Des-gamma carboxyprothrombin can differentiate hepatocellular carcinoma from nonmalignant chronic liver disease in American patients. Hepatology. 2003;37:1114–1121
  21. Sherman M. Alphafetoprotein: an obituary. J Hepatol. 2001;34:603–605

 Conflicts of interest The authors disclose no conflicts of interest.

 Funding Alejandro Forner is partially supported by a grant from the Instituto de Salud Carlos III (PI 05/645). Maria Reig is funded by a grant of the BBVA foundation. This work is partially supported by a grant from the Instituto de Salud Carlos III (PI 08/146) and Acción Transversal en Cáncer. CIBEREHD is funded by the Instituto de Salud Carlos III.

PII: S0016-5085(09)00779-3

doi:10.1053/j.gastro.2009.05.014

Refers to article:

  • Editorial Accompanies ArticleLinking Article with CGH α-Fetoprotein, Des-γ Carboxyprothrombin, and Lectin-Bound α-Fetoprotein in Early Hepatocellular Carcinoma , 10 April 2009

    Jorge A. Marrero, Ziding Feng, Yinghui Wang, Mindie H. Nguyen, Alex S. Befeler, Lewis R. Roberts, K. Rajender Reddy, Denise Harnois, Josep M. Llovet, Daniel Normolle, Jackie Dalhgren, David Chia, Anna S. Lok, Paul D. Wagner, Sudhir Srivastava, Myron Schwartz
    Gastroenterology July 2009 (Vol. 137, Issue 1, Pages 110-118)

Gastroenterology
Volume 137, Issue 1 , Pages 26-29, July 2009

Article Tools