Determinants of alpha-fetoprotein levels in patients with hepatocellular carcinoma: Implications for its clinical use
Edoardo G. Giannini MD, PhD,
Gastroenterology Unit, Department of Internal Medicine, IRCCS-University Hospital San Martino-IST, University of Genoa, Genoa, Italy
Corresponding author: Edoardo G. Giannini, MD, PhD, Gastroenterology Unit, Department of Internal Medicine, IRCCS-University Hospital San Martino-IST, University of Genoa, Viale Benedetto XV, No. 6, 16132, Genoa, Italy; Fax: (011) 39-010-353-8638; email@example.com
Other members of the ITA.LI.CA Group: Mauro Bernardi, Luigi Bolondi, Maurizio Biselli, Paolo Caraceni, Alessandro Cucchetti, Marco Domenicali, Virginia Erroi, Rita Golfieri, Annagiulia Gramenzi, Barbara Lenzi, Donatella Magalotti, Fabio Piscaglia, Matteo Ravaioli, and Laura Venerandi (Department of Medical and Surgical Sciences, Alma Mater Studiorum-University of Bologna, Bologna); Anna Giacomin, Veronica Vanin, Caterina Pozzan, and Gemma Maddalo (Department of Surgical and Gastroenterological Sciences, University of Padua, Padua); Paolo Del Poggio and Stefano Olmi (Operative Surgical Unit, San Marco Polyclinic, Zingonia); Claudia Balsamo, Maria Anna Di Nolfo, and Elena Vavassori (Operative Medical Unit, Bolognini Teaching Hospital, Seriate); Alfredo Alberti, Luisa Benvegnù, Angelo Gatta, and Maurizio Gios (Department of Clinical and Experimental Medicine, University of Padua, Padua); Emanuela Giampalma, Rita Golfieri, Cristina Mosconi, and Matteo Renzulli (Operative Radiology Unit, Department of Digestive Tract Diseases and Internal Medicine, Teaching Hospital-University of Bologna, Bologna); Giulia Bosco (Unit of Internal Medicine and Gastroenterology, Columbus Integrated Complex, Catholic University of Rome, Rome); Paola Roselli (Operative Gastroenterology Unit, Belcolle Hospital, Viterbo); Serena Dell'Isola and Anna Maria Ialungo (Operative Unit of Preventive Medicine, Belcolle Hospital, Viterbo); Vincenzo Savarino, Antonino Picciotto, Domenico Risso, Simona Marenco, and Linda Bruzzone (Gastroenterology Unit, Department of Internal Medicine, University of Genoa, Genoa); Maria Chiaramonte (Gastroenterology Unit, Don Calabria Sacred Heart Hospital, Negrar); Calogero Cammà, Marcello Maida, Arezia Di Martino, and Maria Rosa Barcellona (Gastroenterology Unit, Biomedical Department of Internal and Specialty Medicine, University of Palermo, Palermo); Andrea Mega (Gastroenterology Unit, Bolzano Regional Hospital, Bolzano); Emanuele Rinninella (Unit of Internal Medicine and Gastroenterology, Gemelli Polyclinic, Catholic University of Rome, Rome); Valeria Mismas (Operative Unit for Gastroenterology and Metabolic Diseases, Teaching Hospital-University of Pisa, Pisa); Arianna Lanzi, Giuseppe Francesco Stefanini, Anna Chiara Dall'Aglio, Federica Mirici Cappa, Elga Neri, Paolo Bassi, and Miriam Zanotti (Department of Internal Medicine, Faenza Hospital for the Infirm, Faenza); Elisabetta Biasini and Emanuela Porro (Unit of Infectious Diseases and Hepatology, Teaching Hospital-University of Parma, Parma); Maria Guarino (Gastroenterology Unit, Department of Clinical Medicine and Surgery, Federico II University of Naples, Naples); and Laura Schiadà and Stefano Gemini (Gastroenterology Clinic, Marche Polytechnic University, Ancona).
α-Fetoprotein (AFP) is a biomarker commonly used in the management of patients with hepatocellular carcinoma (HCC), although the possible determinants of its serum levels in these patients have not been adequately explored. For this study, the authors evaluated the relevance of demographic, clinical, and oncologic factors to the presence of elevated AFP levels in large cohort of patients with HCC.
In 4123 patients with HCC who were managed by the Italian Liver Cancer Group, AFP levels were assessed along with their association with demographic, biochemical, clinical, and oncologic characteristics. Patients were subdivided according to the presence of elevated AFP (ie, >10 ng/mL).
AFP levels were elevated in 62.4% of patients with HCC. Multivariate logistic regression analysis indicated that being a woman (odds ratio [OR], 1.497; 95% confidence interval [95%CI], 1.250-1.793; P < .0001), the presence of cirrhosis (OR, 1.538; 95% CI, 1.050-2.254; P = .027), liver disease with viral etiology (OR, 1.900; 95% CI, 1.589-2.272; P < .0001), an elevated alanine aminotransferase level (OR, 1.878; 95% CI, 1.602-2.202; P < .0001), a low albumin level (OR, 1.301; 95% CI, 1.110-1.525; P = .012), an HCC tumor size >2 cm (OR, 1.346; 95% CI, 1.135-2.596; P = .001), multinodular HCC (OR, 1.641; 95% CI, 1.403-1.920; P < .0001), and the presence of vascular invasion (OR, 1.774; 95% CI, 1.361-2.311; P < .0001) were associated independently with elevated levels of AFP. Both the median AFP level and the proportion of patients who had elevated levels increased with decreasing degrees of HCC differentiation (P < .0001).
Hepatocellular carcinoma (HCC) is the third cause of cancer death and the leading cause of mortality among cirrhotic patients. The main risk factor for the development of HCC is the presence of cirrhosis, with an annual incidence of HCC in these patients of 3% to 7%.[2, 3] Detection of HCC at an early stage in patients with preserved liver function is associated with a greater likelihood of being amenable to curative treatment and, thus, with improved patient survival.[4-8] Therefore, surveillance for HCC is currently recommended for patients with compensated, chronic liver disease who have an annual incidence of HCC that triggers a favorable cost-effective surveillance program.[9-11]
α-Fetoprotein (AFP) is a biomarker whose serum concentration increases in most patients with HCC; therefore, it has been used in the surveillance for and diagnosis of HCC in patients with chronic liver disease. The latest Western guidelines for the clinical management of HCC dropped AFP from the screening and surveillance armamentarium, because it was proven to be both not sensitive enough to identify early stage HCC and not specific enough to avoid unnecessary recall procedures.[9-11, 13-23] Nevertheless, despite this compelling evidence, the use of serum AFP in the surveillance of HCC is a “hard-to-die” issue, and its monitoring over time—rather than a single determination—has been proposed to identify patients with chronic liver disease who are most likely to develop HCC.[24, 25] Moreover, it has been suggested that AFP determination may improve the prognostic assessment and treatment of patients with HCC.[26-29]
Among patients who have chronic hepatitis C virus infection without HCC, mild to moderate AFP elevations have been observed more frequently in women, patients with high biochemical activity, and those with advanced histology.[15, 17, 19] However, although previous studies in relatively small or selected populations of patients with HCC reported an association of serum AFP levels with some unfavorable HCC characteristics, the correlations between circulating AFP levels and both patient and tumor characteristics have been thoroughly evaluated in adequately sized populations only among Asian patients.[28, 29]
The objective of the current study of 4123 Western patients with HCC was to explore the association between AFP levels and both patient and tumor characteristics. We hypothesized that assessment of the relation between serum AFP and both patient and tumor characteristics would allow us to identify factors that should be taken into account when evaluating elevated levels of this biomarker and eventually would help clinicians to refine the use of serum AFP in the diagnosis and clinical management of patients with HCC.
MATERIALS AND METHODS
The Italian Liver Cancer (ITA.LI.CA) database currently contains data on 5136 HCC patients who were consecutively diagnosed with HCC from 1987 to 2010 at 18 Italian medical institutions in Italy. These data have been collected prospectively and are updated every 2 years with information on the follow-up of the patients. After data entry by any single center, the consistency of the data set is checked by the group coordinator (F.T.); and, when clarification or additional information is needed, it is resubmitted to each center before statistical evaluation.
For the purpose of this study, we included all patients who had serum AFP levels reported at the time of HCC diagnosis (n = 4123; 80.3%). Patients were analyzed according to the presence of elevated AFP levels (ie, >10 ng/mL).
Liver tests, tests for identifying the etiology of cirrhosis, and AFP levels were determined by conventional methods using commercially available assays. The presence of cirrhosis was assessed by the physician in charge of the patient according to histologic or unequivocal clinical and instrumental evidence. Liver function was clinically evaluated using the Child-Pugh classification.
The diagnosis of HCC was made according to the guidelines published at the time of patient inclusion. Tumor size and disease stage were assessed by radiologic imaging, and the macroscopic types of HCC were classified as uninodular, multinodular, diffuse, and massive. HCC histology was assessed following published criteria and was graded according to the Edmondson and Steiner classification.[32, 33]
Continuous data are expressed as median values and 95% confidence intervals (CIs) of the median, and discrete variables are expressed as absolute and relative frequencies. To compare continuous variables, we used the Mann-Whitney U test, and discrete variables were compared using the chi-square test with Yates correction or the Fisher exact test, as appropriate.
Univariate analyses were carried out to explore the association between elevated AFP levels and demographic, clinical, and oncologic variables. Variables that were associated with P values ≤ .1 in univariate analysis were entered into a multivariate logistic regression analysis. In this analysis, we used the following cutoff values for continuous variables: for albumin, the lower limit of the normal range (ie, 3.5 mg/dL); for bilirubin, the upper limit of the normal range (ie, 1.1 mg/dL); for alanine aminotransferase, the upper limit of normal; for HCC size we selected a 2-cm cutoff value, because this is the target of HCC screening and surveillance, although we also performed further analyses with 3-cm and 5-cm cutoff values. A 2-tailed P value < .05 was considered statistically significant. Statistical analyses were performed using the MedCalc statistical software package (MedCalc Software, Mariakerke, Belgium).
Management of the ITA.LI.CA database conforms to past and current Italian legislation on the privacy, and the current study conformed to the ethical guidelines of the Declaration of Helsinki. Approval for this study was obtained by the institutional review boards of the participating centers.
The main characteristics of the 4123 patients who were included in this study are provided in Table 1. The ratio of men to women was 3:1, and 2600 patients (63.1%) were aged ≥65 years. Chronic infection with hepatitis viruses was the cause of liver disease in 3020 patients (73.2%), and infection with hepatitis C virus and hepatitis B virus infection alone was the cause of 47.4% (n = 1955) and 9.3% (n = 384) of all disease, respectively. The majority of patients had well compensated cirrhosis (Child-Pugh class A: n = 2143; 57.5%) and uninodular HCC (n = 2034; 50.4%), whereas both vascular invasion (n = 599; 14.8%) and extrahepatic spread (n = 119; 3.1%) had low prevalence. Overall, 2574 patients (62.4%) had elevated serum AFP levels (ie, >10 ng/mL).
Table 1. Main Characteristics of the 4123 Patients at Diagnosis of Hepatocellular Carcinoma
No. of Patients (%) or Median [95% CI]
Abbreviations: CI, confidence interval; ULN, upper limit of normal.
This value indicates the number of patients who had available data. When not specified, information was available in all patients.
Serum AFP and Demographic, Clinical, and Oncologic Characteristics
Table 2 lists the main characteristics of patients subdivided according to the AFP cutoff level of 10 ng/mL. In univariate analysis, an elevated AFP was associated with women (27.5% vs 20.4%; P < .0001), higher alanine aminotransferase levels (1.6 vs 1.0 times the upper limit of normal; P < .0001), lower albumin levels (3.5 g/dL vs 3.7 g/dL; P < .0001), higher bilirubin levels (1.30 mg/dL vs 1.20 mg/dL; P < .0001), the presence of cirrhosis (96.7% vs 93.6%; P = .0002), advanced liver disease (Child-Pugh class A: 54.8% vs 61.8%; P < .0001), viral etiology (79.3% vs 68.8%; P < .0001), HCC size (median greatest dimension, 3.2 cm vs 3.0 cm; P < .0001), gross HCC pathology (multinodular HCC: 60.4% vs 44.4%; P < .0001), vascular invasion (18.3% vs 8.8%; P < .0001), and extrahepatic spread (3.6% vs 2.3%; P = .029).
Table 2. Characteristics of Patients Subdivided According to Serum α-Fetoprotein Levels
Serum AFP Level: No. of Patients (%) or Median Value [95% CI]
Multivariate logistic regression analysis (Table 3) indicated that elevated AFP levels were associated with women (P < .0001), the presence of cirrhosis (P = .027), viral etiology of liver disease (P < .0001), any alanine aminotransferase elevation (P < .0001), low albumin levels (P = .012), HCC size >2 cm (P = .001), multinodular HCC (P < .0001), and the presence of vascular invasion (P < .0001). The variables that were significantly associated with elevated AFP levels in multivariate analysis did not change when the HCC cutoff size was set at either 3 cm (odds ratio [OR], 1.513; 95% confidence interval [CI], 1.293-1.771; P < .0001) or 5 cm (OR, 1.758; 95% CI, 1.415-2.184; P < .0001).
According to the variables identified in multivariate analysis, we built some clinical vignettes demonstrating how the probability of having an elevated AFP level in patients with HCC increased from a minimum of 17.5% in a man without cirrhosis who had liver disease of nonviral etiology, normal albumin and alanine aminotransferase levels, and uninodular HCC without vascular invasion to 90.5% in a woman who had cirrhosis of viral etiology, altered alanine aminotransferase levels, low albumin levels, multinodular HCC >2 cm, and vascular invasion (Fig. 1).
Serum AFP Levels and HCC Histology
Data regarding HCC histology were available for 873 patients (21.2%). The HCC histologic type was trabecular in 480 patients (55%), acinar in 125 patients (14.3%), mixed (trabecular/acinar) in 195 patients (22.3%), solid in 48 patients (5.5%), cholangio-HCC in 17 patients (1.9%), fibrolamellar in 5 patients (0.6%), and scirrous in 3 patients (0.3%). Grading of HCC according to the Edmondson-Steiner classification was available in 781 patients (18.9%) and was grade 1 in 281 patients (36%), grade 2 in 302 patients (38.7%), grade 3 in 174 patients (22.3%), and grade 4 in 24 patients (3.1%).
Table 4 lists the median AFP levels and the distribution of patients who had AFP elevations according to HCC histologic type and grade. No specific AFP pattern was associated with the various HCC histologic types (P = .284 for comparison among the trabecular, acinar, mixed, and solid types; P = .108 for the same analysis including the cholangio-HCC type). Instead, AFP levels clearly increased progressively with decreasing HCC differentiation from grade 1 to grade 4. Accordingly, the prevalence of elevated AFP levels increased with decreasing HCC differentiation (grade 1, 47.7%; grade 2, 56.9%, grade 3, 66.7%, grade 4, 83.3%; P < .0001).
Table 4. Serum α-Fetoprotein Levels and Proportion of Patients With Altered Levels in Patients With Various Hepatocellular Carcinoma Histologies and Grades
No. of Patients (%)
Median AFP Level, ng/mL
AFP >10 ng/mL: No. of Patients (%)
Abbreviations: AFP, α-fetoprotein.
Histologic type, n = 873
Grade, n = 781
Serum AFP in Patients With Compensated Cirrhosis of Viral Etiology and a Single, Tiny (≤2 cm) HCC Without Vascular Invasion or Extrahepatic Spread
The identification of a single, tiny HCC in patients who have compensated liver disease is the target of surveillance, because these patients theoretically should be amenable to curative treatment with the best results and, thus, should benefit most from surveillance. Therefore, we selected a subgroup of 295 patients who had well compensated cirrhosis (ie, Child-Pugh class A) of viral etiology and a single, small HCC (ie, ≤2 cm) without vascular invasion or extrahepatic spread. Viral etiology was selected for 2 reasons: 1) it represented the leading cause of HCC, and 2) it allowed us to perform the analysis controlling for 1 additional confounding factor for AFP alteration. Elevated AFP levels were present in 177 patients (59.8%.) and were associated independently with women (OR, 1.759; 95% CI, 1.033-2.995; P = .037) and with elevated alanine aminotransferase levels (OR, 3.951; 95% CI, 2.342-6.665; P < .0001). It is noteworthy that the median size of HCC was similar between women and men (2.0 cm vs 1.8 cm, respectively; P = .372).
Serum AFP has been used as a surveillance test for HCC in patients with chronic liver disease and as a diagnostic tool in patients with focal liver lesions and suspected HCC.[12, 13] Nevertheless, the most recent Western guidelines for the management of HCC have excluded serum AFP from both the surveillance and diagnostic armamentarium of HCC because of its insufficient sensitivity and low specificity—leading to unnecessary recall procedures—as a screening and surveillance test and because of its inadequate specificity as a diagnostic tool.[9-11, 13] Mainly, there are 2 causes of the poor performance of AFP for HCC detection: first, chronic liver disease per se may stimulate AFP production, because elevated AFP levels are associated with both hepatic necro-inflammatory activity and advanced liver disease in patients without HCC; second, in patients with HCC, elevated AFP levels reportedly are associated with advanced tumor stage.[15-19, 28, 29] Nevertheless, to our knowledge, a comprehensive evaluation of the association between AFP levels and both patient and tumor characteristics obtained in a very large cohort of Western patients with HCC has not been carried out. We hypothesized that such an evaluation could identify the principal factors that should be taken into account in interpreting the oncologic meaning of this biomarker and could help refine its use in the management of patients with suspected or known HCC.
The results of our study indicate that, in patients with HCC, elevated AFP levels are associated with women, elevated alanine aminotransferase levels (reflecting hepatic necro-inflammatory activity), low serum albumin (a marker of decreased liver function), and viral etiology of the disease. These “confounding” factors, which are inherent mainly to the underlying liver disease, should be considered when evaluating the oncologic meaning of an elevated AFP level in the workup of patients with suspected or known HCC and when AFP-based decisions concerning patient management are in order.
The association between elevated AFP and aminotransferase levels is well known and is caused by hepatic necro-inflammatory activity and regeneration, which stimulate AFP production by hepatocytes. Therefore, an elevated AFP also may be interpreted as an index of proliferative activity.[34, 35] Our study demonstrated that, in a patient who has HCC with cirrhosis of viral etiology, the presence of altered aminotransferase increased by 15% (in both men and women) the probability of having an elevated AFP level. In clinical practice, this association is especially relevant, in that an AFP elevation in patients who achieve normalization of liver biochemistry after treatment-induced viral clearance should alert physicians, because there is a high probability that it reflects the presence/occurrence of HCC.[36, 37] Although it may be argued that the greater hepatic necro-inflammatory activity and decreased liver function observed in the current study may have been caused by the presence of HCC itself, this hypothesis is quite unlikely, because these findings were independent of tumor characteristics, and most patients had compensated liver disease and small, uninodular HCC without vascular invasion. Furthermore, when we focused the analysis on patients who had well compensated liver disease and tiny (ie, ≤2 cm), uninodular HCC without vascular invasion, an elevated aminotransferase level was the strongest independent determinant (OR, 3.951) of an elevated AFP level.
Another finding of our study that deserves a comment is the association between elevated AFP and women. Indeed, previous studies—albeit with smaller patient numbers—reported this gender effect.[16, 19] In particular, a subanalysis of the Hepatitis C Antiviral Long-Term Treatment Against Cirrhosis (HALT-C) cohort indicated that, compared with men, women more frequently had elevated AFP levels (ie, >20 ng/mL) independent of the presence of HCC. Moreover, in previous studies that included a fraction of our population, we observed that women with HCC had significantly higher serum AFP levels than men, and there was a significant association between women and very high serum AFP levels (ie, >200 ng/mL).[16, 38] This association cannot be considered spurious, because the current large study confirmed that being a woman is independently associated with any elevation in AFP (ie, >10 ng/mL). All in all, for women who had HCC, there was a 4% to 10% greater probability of having elevated AFP compared with men who had HCC, and a consistent trend was confirmed by taking into account the other important confounding factors, such as alterations in liver biochemistry, stage of liver disease, and HCC features (Fig. 1). Although it was not possible to ascertain the role played by hormones in these patients because we did not collect serum samples, it should be pointed out that the median age of the women in our cohort was 71 years, and only 25 of 1023 women (2.4%) were aged ≤50 years, suggesting that the HCC-independent elevation of AFP in women may not be strictly related to sexual hormone production. However, further studies assessing the causes of this gender effect on AFP and the utility of sex-specific cutoff values for this biomarker in the management of patients with suspected or known HCC are warranted.
From an oncologic standpoint, we observed that a large lesion and the presence of multinodular HCCs and vascular invasion were independently associated with elevated AFP levels. These findings underscore the association of elevated AFP levels and advanced tumor stage, which cannot be the target of screening or surveillance procedures and, thus, must be rely on ultrasonographic monitoring of the liver. We observed that a man who had compensated viral cirrhosis, an altered alanine aminotransferase level, and a uninodular HCC <2 cm without vascular invasion—1 of the most likely clinical vignettes of HCC diagnosed during surveillance in the Western world—had only a 53.9% probability of having an elevated AFP level, and that proportion decreased impressively to 17.5% in a similar patient who had liver disease of nonviral etiology, a normal alanine aminotransferase level, and no cirrhosis. These results herald the need for a reassessment of the role of AFP in the changing scenario of HCC risk factors, characterized by a mounting prevalence of tumors developing in patients with nonviral liver disease, such as alcoholic cirrhosis and nonalcoholic fatty liver disease.[7, 39-41] Indeed, less frequently and to a lesser extent, the AFP level is elevated in these patients with HCC but, when elevated, may be more reliable as a decision aid than in patients who have viral liver disease, because it more confidently reflects the presence of tumor.
Finally, a subgroup analysis of the current cohort demonstrated a clear association between elevated AFP levels and a lower grade of HCC differentiation, providing evidence for the possible role of AFP in highlighting tumors with malignant biologic activity and, likely, a worse prognosis. This ability of AFP to indicate tumor aggressiveness supports the suggested prognostic role of this biomarker and explains its relevance in the clinical decision-making process for many settings, including liver transplantation, locoregional treatments, and systemic therapy.[42-44]
In conclusion, the current study demonstrates that, in patients with HCC, an elevated AFP level is associated with women, necro-inflammatory activity, liver disease of viral etiology, decreased liver function, and advanced stage of HCC. All of these factors should be taken into account when evaluating the diagnostic and prognostic role of AFP in patients with suspected or known HCC.