J Cancer Res Clin Oncol (2004) 130: 417–422

DOI 10.1007/s00432-004-0552-0

O R I GI N A L P A P E R

Bo-Heng Zhang Æ Bing-Hui Yang Æ Zhao-You Tang

Randomized controlled trial of screening for hepatocellular carcinoma

Received: 25 September 2003 / Accepted: 28 January 2004 / Published online: 20 March 2004
Ó Springer-Verlag 2004

Abstract Purpose: Screening for hepatocellular carci- Keywords Hepatocellular carcinoma Æ Screening Æ
noma (HCC) has been conducted for over 20 years, but Randomized controlled trial Æ Mortality
there is no conclusive evidence that screening may
reduce HCC mortality. The aim of this study was to
assess the effect of screening on HCC mortality in people
at increased risk. Methods: This study included 18,816 Introduction
people, aged 35–59 years with hepatitis B virus infection
or a history of chronic hepatitis in urban Shanghai, Liver cancer is the third most common cause of death
China. Participants were randomly allocated to a from cancer in the world, with an estimated 548,600
screening (9,373) or control (9,443) group. Controls deaths in the year 2000 (Parkin et al. 2001). Since the
received no screening and continued to use health-care 1990s, HCC has become the second cancer killer in
facilities. Screening group participants were invited to China. Although progress has been made in the man-
have an AFP test and ultrasonography examination agement of symptomatic HCC, there has been little
every 6 months. Screening was stopped in December overall reduction in HCC mortality during the past
1997; by that time screening group participants had been 30 years. Risk factors for HCC include hepatitis B and
offered five to ten times. All participants were followed hepatitis C virus infections, dietary intake of aflatoxins,
up until December 1998. The primary outcome measure and drinking water contamination in the rural area (Yu
was HCC mortality. Results: The screened group 1995). In China, the hepatitis B virus is particularly
completed 58.2 percent of the screening offered. When prevalent, and vaccinating newborn babies against
the screening group was compared to the control hepatitis B is anticipated to control the incidence of
group, the number of HCC was 86 versus 67; sub- HCC in the future. However, early diagnosis before
clinical HCC being 52 (60.5%) versus 0; small HCC development of symptoms may also be helpful in the
39 (45.3%) versus 0; resection achieved 40 (46.5%) control of this disease.
versus 5 (7.5%); 1-, 3,-, and 5-year survival rate The detection of alpha-fetoprotein (AFP) in the
65.9%, 52.6%, 46.4% versus 31.2%, 7.2%, 0, respec- serum of an HCC patient by Tatarinov provided a new
tively. Thirty-two people died from HCC in the clue for the early detection and diagnosis of HCC
screened group versus 54 in the control group, and (Tatarinov 1964). Screening for HCC has been con-
the HCC mortality rate was significantly lower in the ducted since the 1970s. However, there has been no
screened group than in controls, being 83.2/100,000 direct evidence of the efficacy of screening in people at
and 131.5/100,000, respectively, with a mortality rate increased risk in reducing mortality from HCC.
ratio of 0.63 (95%CI 0.41–0.98). Conclusions: Our From 1971 to 1976, nearly two million people were
finding indicated that biannual screening reduced HCC screened with AFP for HCC in Shanghai. Of 300 HCC
mortality by 37%. patients detected, 134 were diagnosed as subclinical
HCC. The 3-year survival rate of patients after resection
in this group was 57.1%. The usefulness of an AFP
screening was not widely accepted on the basis of results
B.-H. Zhang (&) Æ B.-H. Yang Æ Z.-Y. Tang from Africa (Purves 1976). The screening modality has
Liver Cancer Institute of Fudan University, been changed from using AFP alone in people at
Zhongshan Hospital, 180 Feng Lin Road,
Shanghai 200032, China
average risk to the combination of AFP and ultraso-
E-mail: [email protected] nography in people at increased risk, since the 1980s
Fax: +86-21-64037181 (Purves 1976; Shanghai Coordinating Group for
418

Fig. 1 Study profile

Research on Liver Cancer 1979; Mima et al. 1994; Izzo met the criteria of the program. These subjects were recruited into
et al. 1998). It has been demonstrated that periodic our program from January 1993 to December 1995. With a view to
feasibility and potential bias of the study, simple cluster sampling
screening results in a significant shift to an earlier stage, was carried out. Every ‘factor’, ‘enterprise’, or ‘school’ was re-
asymptomatic cancer at diagnosis (Shanghai Coordi- garded as a unit. This ensured that all eligible members of the unit
nating Group for Research on Liver Cancer 1979; were allocated to the same group. These units were randomly
McMahon and London 1991), and a significant increase allocated to a screening (9,757) or no screening (control, 9,443)
group. In the screening group, 384 subjects refused to participate to
in survival with HCC (Tang et al. 1980). the program (Fig. 1). Controls were identified but received no
Since there were no randomized controlled trials of intervention, and continued to use health-care facilities as usual.
screening for HCC, nor adequate evidence of reduction This approach was judged to be ethical at the time of study design.
of mortality, we conducted a randomized controlled trial The study was approved by Fudan Medical School Ethics Com-
to assess the effect of biannual screening with a combi- mittee.
The screening group participants were invited by GPs to have a
nation of AFP and ultrasonography on HCC mortality serum AFP test [ELISA, kits produced by Tianyuan, China (Yang
in the people at increased risk in urban Shanghai, China. et al. 1997)], and a screening ultrasonography examination every
6 months. The cut-off value of AFP was 20 lg/l. An abnormality
detected by ultrasound was a solid lesion in the liver. Individuals
who did not take the screening test had another chance later. Be-
Materials and methods fore the program initiation, laboratory technicians and ultrasound
doctors from primary care centers were trained in Zhong Shan
The Liver Cancer Institute of Fudan University has studied the Hospital of Fudan University.
secondary prevention of HCC since the 1970s. According to In order to minimize the false positive rate, a repeat test was
previous studies (Yang et al. 1987; Yang et al. 1988; Zhang et al. sent to individuals with a positive AFP test or abnormal screening
1995), people aged 35 years to 59 years and with serum evidence of ultrasound. Only those individuals with positive results at the retest
hepatitis B virus (HBV) infection or a history of chronic hepatitis were offered diagnostic evaluation at the institute. The diagnostic
have an increased risk for HCC (Zhang et al. 1995), and were protocol included a history and physical examination, liver func-
eligible for the study. The serum evidence of HBV infection was tion tests, AFP (radioimmunoassay, RIA), and an ultrasonography
defined as any of five markers of HBV (HBsAg, HBsAb, HBcAb, examination by a senior doctor. Computed tomography or mag-
HBeAg, HBeAb) being detectable in serum, except for the condi- netic resonance imaging was checked when necessary. The final
tion when HBsAb is positive and the others are negative. We used diagnoses were reached by biopsy or long-term follow up. Some
an ELISA test to measure the antigens and antibodies of HBV. The participants with positive screening tests went to other hospitals for
HBV test kits were produced by Tianyuan Company, China. Be- the diagnostic evaluation. Detail information on the results of
fore these kits were widely applied, we tested the agreement of examinations was obtained from their medical records by their
results with these kits against those yielded by Abbot kits and GPs. The validity of screening tests was reported elsewhere (Zhang
calculated the kappa. The agreements of five HBV markers and Yang 1999). When AFP and ultrasonography were used in
(HBsAg, HBsAb, HBcAb, HBeAg, HBeAb) were 97.5%, 95%, parallel, the detection rate, false positive, and positive predictive
95%, 90%, and 92.5%, respectively, and the kappas were 0.93, values were 92%, 7.5%, and 3.0%, respectively.
0.80, 0.89, 0.71, and 0.85, respectively. A history of chronic hepa- Screening-group participants who were found to have HCC
titis was defined as patients with abnormalities on serum bio- were treated and transferred to follow-up programs. Individuals
chemical tests lasting for 6 months or more. The most with negative screening tests were invited to repeat screening every
characteristic pattern is an elevation in both alanine and aspartate 6 months. We stopped screening in December 1997; by that time,
aminotransferase (ALT and AST), and also includes an elevation in all participants had been offered screening tests between five to ten
serum bilirubin. In cooperation with the doctors (GP) in primary times.
care centers, we selected more than 300 factories, enterprises, and We obtained information on the development of HCC and
schools in urban Shanghai. According to the medical records in the death in screening-group participants and unscreened controls
primary care centers, we excluded those with a known history of from the GPs in the factories, enterprises, and schools. Because of
HCC, or other malignant diseases, or serious illness. Thus, 19,200 the requirements of the medical insurance system in Shanghai at
 419

that time, all patients had to be registered in primary care centers Table 1 Age and sex of participants at initial screening
before they were referred to other hospitals. We also cross-checked
with the Shanghai Cancer Registry in early 1999 to confirm that no Screening Control
cases of HCC had been missed. (9,373) (9,443)
The staging systems such as TNM need surgical evaluation, but
not all patients received surgical treatment in our program. We Sex
used the HCC staging system of China (Tang 1989). Three disease Male 5,869 5,979
stages were differentiated: stage I (subclinical stage or early stage) Female 3,504 3,464
refers to HCC patients without obvious cancer symptoms and Mean age 42 41
signs; stage II (moderate stage) refers to those between stage I and No. HBsAg+ 6,071 6,027
stage III, i.e., patients with symptoms or signs of HCC, such as No. HBsAg+ & history 2,508 2,644
palpatable mass in the abdomen; stage III (late stage) refers to of hepatitis
those HCC patients with obvious cachexia, jaundice, ascites or No. history of hepatitis 794 772
distant metastases. The diameter of a tumor less than 5 cm is
empirically defined as small HCC.
Interval cases refers to those HCC patients who complied with
screening, but were found to have liver cancer before the next
screening. In fact, these patients were false negative. Non-re-
sponder patients referred to those HCC patients who did not
comply with screening, but whose details were reported by their
GPs, or who were found in the Shanghai Cancer Registry.
The primary outcome measure was mortality from HCC. Using
the world standard population, the incidence and mortality of
HCC were standardized. Rates of death from HCC in the screening
and control groups were compared by a Poisson Model to calculate
the confidence interval (CI). Cumulative survival was calculated by
the life-table method from the date of diagnosis of HCC, censoring
at the date of death or at 31 December 1998. Survival in the two
groups was compared by the log-rank test. Proportions were Fig. 2 Compliance during repeat screening
compared by v2 test.
All analyses were on an intention-to-treat basis. The study was
approved by the Shanghai Medical University Ethical Committee
(currently merged withFudan University Ethical Committee).

Results

Randomization was effective in creating a balance be-

tween two groups in terms of age and sex (Table 1). The
age and sex distributions of the groups did not change
substantially among the study. However, most partici-
pants were younger than 50, and there were more men
than women.
The screened group completed 58.2 percent of the
screening offered (Fig. 2). Screening was carried out a
minimum of one time, and the median was five times.
More than 70 percent of the participants testing positive
were examined at the Liver Cancer Institute of Fudan
University.
During the study, 71 cases of HCC were detected by
screening, three were interval cases, 12 were diagnosed in Fig. 3 Cumulative incidence of HCC in screening and control
the non-responder group, and 67 were diagnosed in the groups
control group. Since most participants in this program
were younger than 50, the standardized incidence of group. Radical surgery was achieved in 40 (46.5%)
HCC was 279.3 per 100,000 in the screening group, and patients in the screening group and only five (7.8%)
267.0 per 100,000 in the control group. The overall patients in the control group (Table 3). In 52 subclinical
standardized incidence of HCC was 268.0 per 100,000. HCCs, 37 patients (71.2%) were radically resected, and
After the first round of screening, the incidences were another two patients were treated by repeated percuta-
similar in the screening and control group (Fig. 3). The neous ethanol injection (PEI) because their tumors were
distribution of the stages of HCC is shown in Table 2. less than 2 cm and they had severe liver cirrhosis.
There were no cases of subclinical HCC in the control Disease-specific survival of individuals with HCC is
group, whereas the proportion of stage III HCC was shown in Table 2. There was a significant survival
significantly higher in the control group than in the advantage for HCC patients in the screening group over
screening group. There were 39 small HCCs in the those in the control group (P<0.01). No patients in the
screened group, whereas there were none in the control control group survived more than 5 years. The survival
420

Table 2 Stage distribution, treatment and survival of patients with

HCC in the screened and control groups (TACE transcatheter
arterial chemoembolization, PEI percutaneous ethanol injection)

Screening group (86) Control group (67)

Stagea
Stage I 52(60.5%) 0(0%)
Stage II 12(13.9%) 25(37.3%)
Stage III 22(25.6%) 42(62.7%)
Small HCC 39(45.3%) 0
Treatment
Resection 40(46.5%) 5(7.5%)
TACE/PEI 28(32.6%) 28(41.8%)
Conservative 18(20.9%) 34(50.7%)
treatment
Survival (%)b
1-year 65.9 31.2
2-year 59.9 7.2
3-year 52.6 7.2
4-year 52.6 0 Fig. 4 Cumulative mortality from HCC in screening and control
5-year 46.4 0 groups
a
v2=61.41, p<0.01
b
Log-rank v2=35.50, p<0.01
and ultrasonography examination every 6 months led to
Table 3 Outcome of screening a reduction of 37% in HCC mortality in individuals
aged 35)59 years with HBV infection or a history of
Screening group Control
group
chronic hepatitis. The result was obtained with 58.2%
compliance to screening. A greater reduction in mor-
Person-years in the study 38,444 41,077 tality might be achieved with higher compliance to
HCC occurrence screening.
No. of cases 86 67 Since the 1970s, a number of screening programs
Total incidence(per 100,000) 223.7 163.1
Rate ratio (95% CI) 1.37(0.99, 1.89)
have been reported, the results ranging from very opti-
mistic to completely pessimistic (Purves 1976; Shanghai
Deaths from HCC
No. of death 32 54
Coordinating Group for Research on Liver Cancer
Total mortality(per 100,000) 83.2 131.5 1979; Yang et al. 1987; Sherman et al. 1995; Chen et al.
Rate ratio (95% CI) 0.63(0.41, 0.98) 1997; McMahon et al. 2000). In 1990, a workshop on
‘‘screening for hepatocellular carcinoma’’ was held in
Alaska to address the questions of whether screening
rates of HCC in different stages are shown in Fig. 3.
should be routinely performed in high-risk populations.
Subclinical cancers had the best prognosis, the 5-year
The questions included: can groups at high risk for
survival reaching 67.8%, while this was only around 30%
development of HCC be identified? Can HCC be
for stage II cancers, and 0% 5-year survival for stage III
detected at an early stage? Is serum AFP elevated in
cancers. The survival rates of stage II and stage III
hepatitis B surface antigen carriers with resectable HCC?
cancers in the screened group and control were similar.
What is the role of ultrasound in early detection of
Vital status was determined for each study partici-
HCC? Can early detection of HCC lead to prolonged
pant followed-up until the end of 1997. Up to December
survival? (McMahon et al. 1991). We designed this
1997, there were 153 HCCs, with 86 deaths from HCC
randomized controlled study to answer these questions.
among the 18,816 participants (Table 3). Although the
A high-risk population for HCC is mainly defined as
total incidence of HCC was virtually identical in two
serum evidence of hepatitis B virus (HBV) or hepatitis C
groups, the total mortality rate from HCC was lower in
virus (HCV) infection. In this study, the incidence of
the screened group (83.2 per 100,000) than in the control
HCC was 268.0/100,000 in all individuals, 359.5/100,000
group (131.5 per 100,000). The rate ratio for mortality
in males, and 88.0/100,000 in females, while the incidence
from HCC was 0.633 (95 percent confidence interval,
of HCC in the natural population in Shanghai was about
0.41–0.98). These results reveal a significant reduction in
20/100,000. Beasley found that the annual incidence of
mortality at 5-year follow-up in the screened group
HCC in HBsAg-positive male carriers in Taiwan was
compared to the control group (Fig. 4).
495/100,000 (Yu 1995). Chen reported the incidence of
HCC with HBsAg carriers is as high as 897.5/100,000
Discussion (Chen et al. 1982). In Qidong County, an epidemic area
of HCC in China, the incidence was 38.1/100,000 in the
In this randomized controlled trial we found that after natural population and 984.6/100,000 in a sub-popula-
5-year follow-up, screening by combined AFP testing tion with a background of liver disease (Zhu et al. 1983).
 421

68% for subclinical to 30.3% for stage II (Fig. 5). The

results for cumulative survival from the time of diag-
noses of HCCs were consistent with the view that earlier
detection by screening results in improved survival. The
survival rate was significantly higher in the patients with
HCC in the screened group than in the control group
(P<0.01). These rates were similar to others as a num-
ber of screening programs have breported that cancers
detected by screening were smaller and more amenable
to potentially curative therapy (Izzo et al. 1998; Mima
et al. 1994; Shanghai Coordinating Group for Research
on Liver Cancer 1979; Sherman et al. 1995; Tang et al.
1980; Tang 1989; Yang et al. 1997). Based on a 16-year
population-based study, McMahon et al. (McMahon
et al. 2000) concluded that screening of HBsAg carriers
with semiannual AFP was effective in detecting most
HCC tumors at a resectable stage and significantly
prolonged survival rates when compared with historical
Fig. 5 Cumulative survival in different stages HCC patients controls in this population. Although there were no
randomized controlled trials of screening, and all results
Cirrhosis is also a risk factor for HCC; however, more reported suffered from lead-time bias, it is clear that
than 90% of cirrhoses are related to HBV infection in screening results in detecting small cancers in asymp-
China. Thus, we did not regard cirrhosis as an inclusion tomatic individuals, and the majority of them are treated
criterion in this study. The incidence of HCC was higher with curative intent.
in the screening group than in the control group which The results from a study conducted in Qidong
may be attributed to the lead-time of screening and over- County showed that screening can detect HCCs at an
diagnosis. From this program, the lead-time of screening earlier stage; however, there was no advantage in
was estimated at about 0.5 years (Yang et al. 1997). reducing mortality from HCC (Chen et al. 1997). This
According to the rough incidence rates (223.7/100,000 vs might be attributed to insufficient early treatment being
163.1/100,000), an over-diagnosis bias of screening seems provided to HCC patients. Only 25% (25 of 100) sub-
to have existed in this study, but it is hard to estimate clinical HCCs detected by screening received proper
how many patients were over-diagnosed. Actually, the treatment, and most patients gave up the treatment in
incidence rates in the two groups were similar after that study (Zhang et al. 1994), whereas 75% subclinical
standardization. HCCs received radical treatment in our study. This may
In the workshop in Alaska, participants concluded be the most important reason for the profound differ-
that AFP is elevated in most HBsAg-positive carriers ence between the two studies. Proper treatment is
with resectable HCC (McMahon et al. 1991). In previous indispensable to HCC patients detected by screening. On
clinical trials, the detection rate using AFP was 60–70%. the other hand, it is questionable whether to screen liver
The sensitivity of AFP was confirmed at 69% in the cancer in those with server liver cirrhosis.
present study (Zhang et al. 1999). In Japan, ultrasonog- In conclusion, biannual screening with combined
raphy surveys have been proven effective in detecting AFP and ultrasound in individuals aged 35–59 years
small HCC (The Liver Cancer Study Group of Japan reduced HCC mortality after 5-year follow-up. Our
1990). In the authors’ institute, ultrasound has been used findings suggest that consideration should be given to a
in addition to AFP screening since the 1980s, and has program of screening using AFP and ultrasound to
been proven to be useful in picking up HCC with nega- reduce HCC mortality in an increased risk population in
tive AFP. However, the median tumor size detected by the developed areas of China.
ultrasonography screening is generally bigger than that
of tumors detected by AFP screening (Yang et al. 1989).
In this study, ultrasonography was proven to be a useful
screening method (Zhang and Yang 1999). References
HCC patients at earlier stages were found signifi-
cantly more often in the screened group versus the Chen DS, Sheu JC, Sung JL, Lai MY, Lee CS, Su CT (1982) Small
hepatocellular carcinoma: a clinicopathological study in thir-
control. There were twice as many stage III cancers in teen patients. Gastroenterology 83:1109–1119
the control group as in the screening group, being 22 Chen JG, Lu JH, Zhang BC, Chen QG, Hui ZX, Shen QJ (1997)
(25.6%) vs 45 (62.9%). The detection of cancer before its Study on effect of screening on mortality of primary cancer of
development into a stage III cancer had a profound liver. Chin J Public Health 16:341–343
Izzo F, Cremona F, Ruffolo F, Palaia R, Parisi V, Curley SA
effect on mortality. No patients with stage III cancers (1998) Outcome of 67 patients with hepatocellular cancer de-
survived more than 5 years, whereas for patients with tected during screening of 1125 patients with chronic hepatitis.
cancer in the earlier stages, 5-year survival ranged from Ann Surg 227:513–518
422

McMahon BJ, London T (1991) Workshop on screening for Tatarinov Y (1964) Detection of embryo-specific-globulin in the
hepatocellular carcinoma. J Natl Cancer Inst 83:916–919 blood sera of patients with primary liver tumors. Vopr Med
McMahon BJ, Bulkow L, Harpster A, Snowball M, Lanier A, Khim 10:90–91
Sacco F, Dunaway E, Williams J (2000) Screening for hepato- The Liver Cancer Study Group of Japan (1990) Primary liver
cellular carcinoma in Alaska natives infected with chronic cancer in Japan. Clinicopathologic features and results of sur-
hepatitis B: a 16-year population-based study. Hepatology gical treatment. Ann Surg 211:277–287
32:842–846 Yang B, Zhang B, Yu Y, Wang W, Shen Y, Zhang A, Xiang Y, Xu
Mima S, Sekiya C, Kanagawa H, Kohyama H, Gotoh K, Mizuo H Z (1997) Prospective study of early detection for primary liver
(1994) Mass screening for hepatocellular carcinoma: experience cancer. J Cancer Res Clin Oncol 123:357–360
in Hokkaido, Japan. J Gastroenterol Hepatol 9:361–365 Yang BH, Tang ZY (1988) HCC detected in mass screening from
Parkin DM, Bray F, Ferlay J, Pisani P (2001) Estimating the world high risk population. Chin J Digest 8:130–133
cancer burden: Globocan 2000. Int J Cancer 94:153–156 Yang BH, Tang ZY (1989) The value of real-time ultrasonography
Purves LR (1976) Alpha-fetoprotein and the diagnosis of liver cell in massive screening for primary liver cancer. Tumor 9:117–118
cancer. In: Cameron HM, Linsell DA, Warwick GP (eds) Liver Yang BH, Liu KD, Tang ZY (1987) Mass survey for hepatocellular
cell cancer. Elsevier, Amsterdam, The Netherlands, pp 61–80 carcinoma in high risk population. Tumor 7:82–83
Shanghai Coordinating Group for Research on Liver Cancer (1979) Yu SZ (1995) Primary prevention of hepatocellular carcinoma.
Diagnosis and treatment of primary hepatocellular carcinoma in J Gastroenterol Hepatol 10:674–682
early stage—report of 134 cases. Chin J Med 92:801–806 Zhang B, Yang B (1999) Combined alpha-fetoprotein testing and
Sherman M, Pelterkiam KM, Lee C (1995) Screening for hepato- ultrasonography as a screening test for primary liver cancer.
cellular carcinoma in chronic carriers of hepatitis B virus: J Med Screen 6:108–110
incidence and prevalence of hepatocellular carcinoma in a Zhang BC, Wang MR, Chen JG, Jiang RH, Chen QG, Hui ZG
North American urban population. Hepatology 22:432–438 (1994) Analysis of patients with primary liver cancer detected
Tang ZY, Yang BH, Tang CL (1980) Evaluation of population (1) at mass screening and (2) during follow-up period in a high
screening for hepatocellular carcinoma. Chin Med J 93:795–799 risk population. Chin J Clin Oncol 21:489–491
Tang ZY (1989) Efforts in the past decades to improve the ultimate Zhang B, Yang B, Yu Z (1995) The study of high-risk population
outcome of primary liver cancer. In: Tang ZY, Wu MC, Xia SS for primary liver cancer. Tumor 15:80–82
(eds) Primary liver cancer. Springer, Berlin Heidelberg New Zhu YR, Lu XB (1983) Clinical implication of AFP serosurvey.
York, pp 469–481 Chin J Oncol 5:38–40