Institute of Digestive Disease, The Chinese University of Hong Kong, Hong Kong SAR, China
Department of Medicine and Therapeutics, The Chinese University of Hong Kong, Hong Kong SAR, China
Address reprint requests to: Vincent W.S. Wong, M.D., Department of Medicine and Therapeutics, 9/F Prince of Wales Hospital, 30-32 Ngan Shing Street, Shatin, Hong Kong. E-mail: email@example.com; fax: 852-2637-3852.
Potential conflict of interest: Grace L.H. Wong has served as an advisory committee member for Otsuka and Gilead; she is also on the speakers' bureau for Echosens, Furui, and Otsuka. Henry L.Y. Chan is a consultant for Abbott, Bristol-Myers Squibb, Furui, Gilead, Merck, Novartis, and Roche, has received honoraria for lecturing for Abbott, Bristol-Myers Squibb, Echosens, Gilead, GlaxoSmithKline, Merck, Novartis, and Roche, and has received an unrestricted grant from Roche for hepatitis B research. He is on the speakers' bureau for Abbott, Bristol-Myers Squibb, Echosens, Gilead, GlaxoSmithKline, Merck, Novartis, and Roche. Vincent W.S. Wong has served as an advisory committee member for Roche, Novartis, Gilead, and Otsuka; he is also on the speakers' bureau for Bristol-Myers Squibb, Roche, Novartis, Abbott Diagnostics, and Echosens.
This study was funded, in part, by the Direct Grant of The Chinese University of Hong Kong (project reference no.: 2041703; to G.L.H.W.).
Alpha-fetoprotein (AFP) is the most widely used biomarker for hepatocellular carcinoma (HCC) surveillance, which is criticized as neither sensitive nor specific in active hepatitis and liver cirrhosis. The aim of this study was to determine the performance of AFP as a tumor marker for HCC in entecavir-treated patients with chronic hepatitis B (CHB). This was a retrospective-prospective cohort study of 1,531 entecavir-treated patients under regular HCC surveillance with AFP and ultrasonography. Mean age was 52 ± 12 years; 1,099 (72%) patients were male and 332 (21.7%) had clinical evidence of cirrhosis. At a mean follow-up of 51 ± 13 months, 57 (2.9%) patients developed HCC (median size: 3.3 cm). AFP fluctuated with alanine aminotransferase (ALT) and peaked at the time of starting entecavir, then gradually decreased after. AFP started to increase 6 months before the diagnosis of HCC. The receiver operator characteristic curve (AUROC) of AFP was highest at the time of HCC diagnosis (0.85; 95% confidence interval [CI]: 0.73-0.98) and remained satisfactory at 3 (0.82; 95% CI: 0.73-0.91) and 6 months (0.79; 95% CI: 0.69-0.89) before the diagnosis. Using the conventional AFP cut-off (20 μg/L) at month 0, the sensitivity and specificity to diagnose HCC were 38.6% and 98.9%, respectively. Adopting the lower cut-off value (6 μg/L) of AFP level at month 0, sensitivity was increased to 80.7%, whereas specificity was decreased to 80.4%. Conclusion: On-treatment AFP is a specific tumor marker for HCC in CHB patients receiving entecavir therapy. Adopting a lower cut-off value of AFP level at 6 μg/L would significantly increase the sensitivity for HCC detection. (Hepatology 2014;59:986–995)
area under the receiver operating characteristic curves
chronic hepatitis B
chronic hepatitis C
hepatitis B e antigen
hepatitis B surface antigen
hepatitis B virus
positive likelihood ratio
negative likelihood ratio
negative predictive value
polymerase chain reaction
positive predictive value
receiver operating characteristic
upper limit of normal
Hepatocellular carcinoma (HCC) is a common malignancy in Asia as a result of the endemics of chronic hepatitis B virus (HBV) infection. Risk of HCC in patients with chronic hepatitis B (CHB) can be estimated with prediction scores in either treatment-naïve patients[2-4] or patients receiving antiviral therapy. Because antiviral therapy, such as lamivudine and entecavir, reduces, but does not eliminate, the risk of HCC,[6, 7] cancer surveillance remains indispensable in treated patients who are at high risk.[8, 9]
Alpha-fetoprotein (AFP) is the most widely used biomarker for HCC surveillance. However, at least one third of HCCs do not secrete AFP, such that the sensitivity for HCC detection is generally low (20%-65%). Furthermore, AFP levels are sometimes elevated in patients with active hepatitis and liver cirrhosis who do not have HCC. The low sensitivity and specificity of AFP in detecting early HCC has led to the American Association for the Study of Liver Disease recommendation to use ultrasonography (USG) alone (without AFP) for HCC surveillance. However, the calls for abandoning AFP in the surveillance strategy was considered premature by some experts,[8, 13] given the already low rates of surveillance in community practice or areas in which USG is not readily and timely available.
Most of the currently available data concerning AFP in CHB came from treatment-naïve patients. Patients with active hepatitis and liver cirrhosis have a higher risk of HCC, and these patients are mostly on antiviral therapy. Because patients on antiviral therapy rarely have hepatitis flares, the falsely high AFP levels should be a lesser concern, compared to treatment-naïve patients. Therefore the performance of AFP in CHB patients receiving antiviral therapy should be reexamined. In this large-scale, real-life cohort study, we aimed to determine the performance of AFP as a tumor marker for HCC in entecavir-treated CHB patients. The dynamic changes of on-treatment AFP were investigated to define the optimal timing and cut-off value of AFP level to detect HCC.
Patients and Methods
This was a retrospective-prospective cohort study.[5, 7] The entecavir cohort was composed of consecutive CHB patients who had received entecavir (0.5 mg) daily for at least 12 months in the hepatitis clinics, Prince of Wales Hospital, from December 2005 to March 2013. Patients who received entecavir before October 2009 were retrospectively identified from the HBV DNA record, and were recruited into the prospective follow-up study. All patients newly started on entecavir after October 2009 were recruited into the longitudinal study in a prospective manner. A historical control cohort of 424 treatment-naïve patients was recruited from December 1997 to July 2000 from the hepatitis clinic at the same hospital. These patients underwent routine clinical care until the mid to late 2000s, whenantiviral treatments were not readily available or reimbursable.[2, 14] All patients had positive hepatitis B surface antigen (HBsAg) for at least 6 months at recruitment. Patients suffering from chronic hepatitis C (CHC), preexisting HCC, or HCC diagnosed within the first year of entecavir treatment were excluded. This study was approved by the clinical research ethics committee of the Chinese University of Hong Kong. All patients provided informed written consent.
Clinical and Laboratory Evaluation
At baseline, that is, when entecavir was started, patients received an evaluation, including a full medical history, physical examination, hematological and biochemical parameters, HBsAg, hepatitis B e antigen (HBeAg) and antibody (anti-HBe), HBV DNA, and trans-abdominal USG. HBV DNA, which was measured by Taqman real-time polymerase chain reaction (PCR) assay validated against the EUROHEP standard with a linear range of detection from 20 to 2 × 108 IU/mL. HBsAg was quantified by Architect HBsAg QT (Abbott Laboratories, Abbott Park, IL), with a 1:500 autodilution, according to the manufacturer's instruction. The total range of detection was 0.05 to 124,950 IU/mL.
The patients were followed up once every 1-6 months, depending on the clinical condition. Liver biochemistry and AFP were checked every visit. Patients with elevated AFP would be offered liver imaging (e.g., trans-abdominal USG and/or triphasic computed tomography [CT]) within 3 months, or had AFP rechecked together with liver biochemistry in the presence of elevated alanine amontransferase (ALT). HBV DNA was checked every 6-12 months; maintained virologic response was defined as undetectable serum HBV DNA until the last visit. Trans-abdominal USG was performed every 1-2 years for surveillance of HCC, or more frequently if AFP rose above 20 μg/L. Liver cirrhosis was defined as shrunken small liver with nodular surface noted on liver imaging and clinical features of portal hypertension (e.g., ascites, splenomegaly, and varices).
Diagnosis of HCC
The diagnosis of HCC was established based on histopathological confirmation, detection of a positive lesion with at least two imaging techniques (trans-abdominal USG, triphasic CT, magnetic resonance imaging, or hepatic angiogram), or detection with one imaging technique coupled with an AFP concentration greater than 400 μg/L.
On-Treatment AFP Level and Increase
On-treatment AFP level and increase were investigated in this study. AFP increase from nadir to peak was defined as the difference between the lowest and highest AFP levels. We determined the performance of AFP level and increase in detection of HCC during a 12-month period before the diagnosis of HCC. All HCC cases diagnosed after at least 12 months of entecavir therapy were included. AFP at months −12, −9, −6, −3, and 0 (time of HCC diagnosis) from HCC cases and at corresponding time points from non-HCC cases were analyzed. Time 0 for the non-HCC cases corresponded to the time point 12 months preceding the last imaging to confirm the absence of HCC to avoid subclinical HCC at time 0.
Statistical analysis was performed by the Statistical Package for Social Science (version 20.0; SPSS, Inc., Chicago, IL). Continuous variables were expressed in mean ± standard deviation (SD) or median (interquartile range [IQR]), as appropriate. Qualitative and quantitative differences between subgroups were analyzed using chi-square or Fisher's exact tests for categorical parameters and the Student t or Mann-Whitney's tests for continuous parameters, as appropriate. AFP level and increase at specific time points in the 12 months before the diagnosis of HCC were analyzed with Mann-Whitney's test. For analyses of the performance of AFP in differentiating HCC cases from controls, areas under the receiver operator characteristic curve (AUROC) of two parameters were calculated. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive (LR+) and negative likelihood ratio (LR−), and diagnostic odds ratio (OR) of two cut-off values of AFP levels and increases to detect HCC were reported. Kaplan-Meier's method was used to estimate the cumulative probability of survival in patients with HCC. The log-rank test was used to compare time-to-event curves between patient entecavir and historical control cohorts. All statistical tests were two-sided. Statistical significance was taken as P < 0.05.
Patient Characteristics and Incidence of HCC
During the study period, 1,531 patients fulfilled the inclusion criteria in the entecavir cohort. They had been followed for 51 ± 13 months. The demographic, virologic and clinical characteristics of the patients are summarized in Table 1. Their mean age was 52 ± 12 years, and they were predominantly male (72%) and 332 (22%) patients had cirrhosis. Fifty-seven patients in this cohort developed HCC during the follow-up. The median (range) time point of HCC was at month 24 (range, 12-63) from entecavir therapy. The cumulative incidence rates of HCC at 3 and 5 years were 2.9% (95% confidence interval [CI]: 2.4-3.4) and 4.9% (95% CI 3.9-5.8), respectively. Compared to the non-HCC counterparts, HCC cases were older (59 ± 8 vs. 52 ± 11 years; P < 0.001), had higher baseline AFP levels (21 [range, 3-127] vs. 6 [range, 2-18] μg/L) and more likely to have cirrhosis (68% vs. 20%; P < 0.001). The control cohort included 424 treatment-naïve CHB patients followed for 114 ± 31 months; 53 developed HCC during the follow-up period. The cumulative incidence rates of HCC at 3 and 5 years were 3.7% (95% CI: 2.8-4.6) and 6.5% (95% CI: 5.3-7.7), respectively.
Table 1. Clinical Characteristics of Patients in the Entecavir Cohort
In the entecavir cohort, 31 of 57 patients (54.4%) with HCC were first presented as rising AFP, which triggered subsequent imaging to confirm the diagnosis of HCC. On the other hand, 26 of 57 patients (45.6%) had their HCC diagnosed from screening USG arranged independent of AFP level. The median (IQR) size of HCC at diagnosis was 3.3 cm (range, 2.8-4.2). Most of the patients (50 of 57; 88%) presented with single HCC, whereas 7 had 2-4 tumors at time of diagnosis. Forty-four (77%) patients received treatments for HCC with curative intent; 25 (44%) underwent surgical resection and 19 (33%) received local ablative therapy, namely, radiofrequency ablation (15 patients), microwave ablation (1 patient), and lipiodol-ethanol mixture injection (3 patients). Seven patients (12%) were treated with transarterial chemoembolization; 2 received sorafenib, whereas 2 were provided with supportive care. The median (IQR) survival of these 57 patients was 22 months (range, 14-28). Median tumor size (3.1 vs. 3.4 cm; P = 0.75), proportion of patients with single tumor (84% vs. 92%; P = 0.44), and median survival (21 vs. 22 months; P = 0.91) were similar in patients first diagnosed by AFP and imaging.
In the control cohort, the median (IQR) size of HCC at diagnosis was 4.7 cm (range, 3.2-6.4). Almost two thirds of patients (34 of 53; 64%) presented with single tumor, but only 24 (45%) received curative treatments for HCC. Their median (IQR) survival was 18 months (IQR, 8-27). Patients with HCC had better survival in the entecavir cohort, compared to the historical control cohort (P = 0.019).
Impact of Entecavir on AFP Levels
AFP levels fluctuated with ALT levels before entecavir therapy. Both parameters peaked at the time of starting entecavir and then decreased gradually after. Median ALT level was 125 IU/L at baseline, which was decreased to 42 IU/L at month 3 (P = 0.002) and subsequently stabilized between 24 and 31 IU/L from months 6-24 (all P values <0.001, when compared to baseline). The median AFP level was 6 μg/L at month 0, which was decreased to 4 μg/L at month 3 (P = 0.12), and subsequently stabilized at 3 μg/L from month 6 to 24 (P values ranged from 0.02 to 0.07, when compared to baseline; Fig. 1A).
Changes of AFP Before HCC Diagnosis
Figure 1B displays AFP levels in HCC and non-HCC cases at months −12, −9, −6, −3, and 0 (time of HCC diagnosis). AFP levels increased from 10 μg/L (range, 6-29) at month −12 to 81 μg/L (range, 13-316) at month 0 in the HCC cases, but remained unchanged in non-HCC patients: 4 μg/L (range, 2-5) at month −12 and 4 μg/L (range, 2-4) at month 0, respectively. ALT and HBV DNA levels were comparable in two groups of patients at all these time points.
Figure 1C displays AFP increases from nadir to peak in the HCC and non-HCC cases at different time points before HCC diagnosis. AFP increases were more obvious starting from month −3 (11 μg/L; range, 6-29), then more dramatic at month 0 (16 μg/L; range, 1-298) in the HCC cases. There were only minimal changes in non-HCC cases.
Performance of AFP Level and Increase for HCC
To evaluate the AFP changes shortly before the diagnosis of HCC, we further studied the AFP levels 3 and 6 months before confirming the presence of HCC (i.e., months −3 and −6). Figure 2 illustrates the receiver operating characteristic (ROC) curves at different time points before the diagnosis of HCC; the AUROCs are shown in Table 2. The AUROC of AFP level was higher than that for AFP increase at all time points. The AUROC of AFP level was highest at month 0 (0.85) and remained satisfactory at months −3 (0.82) and −6 (0.79). The AUROC of AFP increase was also highest at month 0 (0.77), but became less satisfactory at other time points (0.58-0.69).
Table 2. AUROC of AFP Value Increase at Different Time Points From Entecavir Commencement for HCC
AFP Increase From Nadir to Peak
Time From HCC Diagnosis
Table 3 shows the sensitivity and specificity of AFP level and increase in differentiating HCC cases from non-HCC patients at months 0, −3, and −6. Two cut-off values (6 and 20 μg/L) were chosen from both parameters; 6 μg/L was chosen because the sum of sensitivity and specificity was highest at this cut-off value. Using the conventional AFP cut-off (20 μg/L) at month 0, the sensitivity and specificity to diagnose HCC were 38.6% and 98.9%, respectively. Adopting the lower cut-off value (6 μg/L) of AFP level at month 0, sensitivity was increased to 80.7%, whereas the specificity was decreased to 80.4%. For AFP increase at the lower cut-off value of 5 μg/L, sensitivity and specificity was 79.0% and 80.9%, respectively, which became 42.1% and 95.9%, respectively, if the cut-off AFP increase was 15 μg/L. If combining two AFP parameters together by adopting an “or” approach, the performance was similar as AFP level alone at lower cut-off values (Table 3). Performances of all cut-off values were still satisfactory at months −3 and −6 (Supporting Table 1). Eight hundred and seven patients (52.7%) had liver imaging performed at intervals of ≤1 year. A sensitivity analysis of the performance of AFP in this subset of patients revealed that the performance of AFP was very comparable to the entire cohort (Supporting Table 2).
Table 3. Performance of AFP Level and Increase From Nadir to Peak in Differentiating HCC Cases From Controls
Table 4 shows the sensitivity and specificity of AFP level at 6 μg/L in differentiating HCC cases from non-HCC patients at months 0, −3, and −6 in the historical control cohort. Sensitivity was slightly higher than those in the entecavir cohort, whereas specificity was significantly lower at all time points (details shown in Supporting Table 3).
Table 4. Performance of AFP Level at 6 μg/L in Differentiating HCC Cases From Controls at Different Baseline HBV DNA Levels in the Historical Control Cohort
Historical Control Cohort
Baseline HBV DNA Level Time From HCC Diagnosis
Entecavir Cohort (N = 1,531)
Historical Control Cohort (N = 424)
<2,000 IU/mL (N = 116)
≥2,000 IU/mL (N = 308)
Abbreviation: NA, not applicable.
HBV DNA Level and Performance of AFP
Undetectable serum HBV DNA was achieved in 954 (62.3%), 1,124 (73.4%), 1,167 (76.2%), and 1,240 (81.0%) of patients at months 6, 12, 24, and 36, respectively. At the last visit, 1,249 patients (81.6%) achieved maintained virologic response. After receiving at least 12 months of entecavir therapy, 394 patients had detectable HBV DNA at 1,773 time points; at 117 of these time points, HBV DNA ≥2,000 IU/mL. ALT was above the upper limit of normal (ULN) at 134 and 35 time points when HBV DNA was detectable and >2,000 IU/mL, respectively. Sensitivity and specificity of AFP at 6 ng/mL was 77.6% and 79.6%, respectively, for those with detectable HBV DNA and 81.3% and 76.8%, respectively, for those with HBV DNA ≥2,000 IU/mL. The corresponding figures in the subgroup of patients with elevated ALT were 84.6% and 62.8% and 75.0% and 71.0% in those with detectable HBV DNA and ≥2,000 IU/mL respectively.
The effect of baseline HBV DNA on the performance of AFP in the historical control cohort was also investigated. Eight of one hundred and sixteen patients with baseline HBV DNA <2,000 IU/mL developed HCC, whereas 45 of the 308 patients with baseline HBV DNA ≥2,000 IU/mL developed HCC. Among patients with baseline HBV DNA <2,000 IU/mL, the performance of AFP at 6 μg/L was comparable to that in the entecavir cohort, though specificity was generally lower. On the other hand, the performance in the subgroup of patients with HBV DNA ≥2,000 IU/mL was less satisfactory, because specificity was significantly lower in all time points (Table 4).
To our knowledge, this is the first large-scale, real-life cohort study to assess the performance of on-treatment AFP in CHB patients. In contrast to the criticism that AFP is neither sensitive nor specific, the current study has provided important evidence that AFP is specific to establish the diagnosis of HCC in patients receiving entecavir, when compared to treatment-naïve patients. This is because the falsely elevated AFP in the setting of active hepatitis would be minimized with antiviral therapy. A lower cut-off value of AFP level at 6 μg/L would be rather sensitive, whereas the conventional cut-off value of 20 μg/L was very specific to detect HCC in patients receiving antiviral therapy. We also demonstrated that HBV DNA level did not affect the performance of AFP, as long as the ALT level was normal.
Table 5 summarizes the performance of AFP in differentiating HCC from non-HCC cases in CHB patients from the literature and the current study. This table includes studies that mentioned the proportion of CHB patients included. The performance of AFP was once found impressive in two cohort studies, because sensitivity was as high as 100% and specificity close to 90%,[17, 18] which made AFP a tumor marker for HCC in CHB.[19, 20] However, subsequent studies demonstrated diversified results.[21-34] Sensitivity of AFP level at 20 μg/L could be as low as 39%, whereas specificity was only 50% at a cut-off value of 25 μg/L. In fact, the performance of AFP was found to be inferior in HBsAg-positive patients, compared to HBsAg-negative patients, because elevated AFP associated with exacerbation of hepatitis may lead to false-positive results and low specificity. The difference in study design also contributed to the diversified performance of AFP. Because case-control studies tend to include cases at the extreme ends of the spectrum (e.g., advanced HCC vs. normal), the performance of a biomarker usually appears better than that in cohort studies. The use of a screening cohort of our study was, in fact, more reflective of the real-life situation.
Table 5. Performance of AFP in Differentiating HCC Cases From Non-HCC Cases in Treatment-Naïve Patients From Previous Studies and in Patients Receiving Antiviral Therapy From the Current Study
We demonstrated that the AFP level and the increase started to obviously increase at month −6 from the diagnosis of HCC. Because the median size of HCC if our study was only 3.3 cm, AFP might provide a chance of early HCC detection before USG can pick up the lesion, particularly in cirrhotic liver with the appearance of nodularity on imaging. This echoed the findings of a recent nested case-control study of CHC patients, in which three AFP metrics (i.e., most recent level, SD in the levels, and velocity) were suggested to associate with HCC. In fact, the SD of all AFP values recorded may not be easily computed in a clinic setting; AFP increase or velocity may be more easily applied clinically. We also explored the performance of SD of AFP in the current study, which did not perform as well as the AFP level or increase (see Supporting Table 4).
The importance of a good HCC surveillance program is reflected by the fact that the 5-year overall survival of HCC patients was only 3%-5% at the end of the last decade, despite all the advances in HCC treatment. This was partly contributed by the absence of an effective method for early diagnosis or the failure of surveillance process. Less than half of HCC cases were suitable for potentially curative treatments at the time of diagnosis. In the real world, adherence to HCC surveillance depends on a multitude of patient, physician, and system-level factors. A simple, widely available serum tumor marker will help to maintain the consistency and improves the uptake of surveillance program. We believe AFP will be particularly useful among family physicians who are taking care of a large proportion of CHB patients in the community. Nevertheless, in view of the suboptimal sensitivity, even at the low cut-off of 6 μg/L, AFP should always be coupled with USG in the surveillance program.
There are several implications concerning the clinical application of AFP as part of the HCC surveillance program in CHB patients receiving antiviral therapy. This should be applicable to all antiviral therapy, because previous studies demonstrated comparable effect of various antiviral therapy on ALT normalization and reduction of various HCCs. Furthermore, the new cut-off value of AFP may also be applicable to patients with inactive disease, that is, those with low viral load (<2,000 IU/mL) and normal ALT level. Because AFP level has a high specificity and the NPV, AFP level should be adopted as the first-line screening tool at the community level. A lower cut-off value (e.g., 6 μg/L) should be used to initiate further imaging confirmation of HCC in patients receiving antiviral therapy without the confounding effect of elevated ALT. Regular USG should be provided at the same time to those at risk (e.g., patients with liver cirrhosis), or those with elevated AFP levels. Patients with raised or rising AFP should undergo more frequent follow-up imaging (e.g., every 3 months), even if the index imaging is negative, to aid an early diagnosis of HCC.
Our study has the strength of a real-life screening cohort of large sample size and long follow-up, which increased the statistical power and reliability of the results. It also allowed the analysis of data up to 12 months before the diagnosis of HCC. Nonetheless, our study also has a few limitations. First, the relative small number of HCC patients might have limited the assessment of the sensitivity of AFP with wide CIs, even in such a large cohort. But, the incidence of HCC would be expected to be low, because entecavir therapy is proven to reduce the risk of HCC. Second, the retrospective nature of certain clinical information might be biased by incomplete data collection. Fortunately, the compliance of AFP was high because it had been a routine test in almost all clinic visits. Furthermore, patients recruited after October 2009 had all the clinical information prospectively collected, which also minimized the chance of incomplete data collection. Third, there might be a cohort effect on patient outcomes, because the shorter median survival in the control group (Fig. 3) might be influenced by availability of treatment modalities for HCC, which have changed over time. Fourth, the variability in USG monitoring in the entecavir cohort might have affected the time intervals for the analysis of AFP change as a predictor of HCC.
In conclusion, on-treatment AFP is a specific tumor marker for HCC in CHB patients receiving entecavir therapy. AFP level starts to increase approximately 6 months before the diagnosis of HCC. Adopting a lower cut-off value of AFP level at 6 μg/L would increase the sensitivity significantly for HCC detection. Elevated AFP to >20 μg/L has a very high specificity for HCC in this clinical setting.