Hepatocellular carcinoma (HCC) is one of the most common malignancies in the world, especially in sub-Saharan Africa and Southeast Asia. Since 1984, it has been the leading cause of cancer death in Taiwan,1 accounting for approximately 7,000 death annually. In addition, around 8,000 new HCC cases are diagnosed each year.2 High-risk groups for HCC include patients chronically infected with hepatitis B virus (HBV) or hepatitis C virus (HCV), liver cirrhosis patients, and people with a family history of HCC. Chronic HBV and HCV infections are the 2 major etiologies of HCC in Taiwan.3
About 15–20% of the general population in Taiwan are chronic carriers of hepatitis B surface antigen (HBsAg).4 HBV carriers have a 20- to 98-fold increased risk of HCC,5, 6 especially among the hepatitis B e antigen (HBeAg)-positive group.7 A recent prospective study has further confirmed that positivity for HBeAg is associated with an increased risk of HCC.8 To combat HBV infection in Taiwan, a mass immunization program has been launched on July 1, 1984, aiming first at the prevention of perinatal mother-to-infant, chronic HBV infection.9 This nationwide vaccination program has markedly decreased the HBV carrier rate10 and childhood HBV-related HCCs (B-HCCs).11 An 80–85% decrease in HCC among Taiwanese adults up to 3–4 decades later is anticipated.12
The prevalence of anti-HCV antibody (anti-HCV) is 2–4% among the general population of Taiwan,13 while it ranges from 0.95 to 2.2% in Taiwanese blood donors.13, 14 The anti-HCV prevalence is approximately 60% in HBsAg-negative HCC patients.13 Reportedly, of the HCC cases in Taiwan as a whole, about 70–80% is related to HBV infection and 10–30% is related to HCV infection.5, 15, 16, 17, 18 Thus, HBV is the leading cause of HCC, while HCV is the second cause of HCC in Taiwan. However, in geographic areas with high prevalences of HCV infection,19, 18 HCV may be the main cause of HCC.19 This suggests that etiologies of HCC vary among different geographic areas of Taiwan.
Although previous studies have provided some clues on the etiologic and geographic variation of HCC across Taiwan, those studies were small and involved few geographically different areas. To gain better control of HCC in Taiwan, it is important to know the secular trends and geographic variation in HBV-related HCCs and HCV-related HCCs. Thus, we conducted this retrospective multicenter study.
B-HCC, HBV-related HCC; B+C-HCC, HBV plus HCV-related HCC; C-HCC, HCV-related HCC; HBeAg, Hepatitis B e antigen; HBsAg, Hepatitis B surface antigen; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; NBNC-HCC, non-HBV-non-HCV-related HCC; SMR, standardized mortality ratio.
Patients and methods
Patients and data collection
This is a retrospective multicenters study. A total of 8 medical centers participated in this study, including National Taiwan University Hospital (2,200 beds), Taipei Veteran General Hospital (2,700 beds) in northern Taiwan; Changhua Christian Hospital (1,500 beds), China Medical University Hospital (1,500 beds), Taichung Veteran General Hospital (1,200 beds) in central Taiwan; National Cheng Kung University Hospital (1,100 beds), Kaohsiung Chang Gung Memorial Hospital (2,100 beds) in southern Taiwan; and Buddhist Tzu Chi General Hospital (850 beds) in eastern Taiwan (Fig. 1). Since large majority of HCC patients throughout Taiwan is diagnosed and treated in the medical centers and these 8 hospitals cover most of the population in the respective area, the patients enrolled in our study could represent the patient population in Taiwan. Patient records from 1981–2001 with a diagnosis of ICD-9 code = 155 were retrieved from the computer database of each hospital. The medical chart was then reviewed by hepatologists and experienced and well-trained specialized nurses. The ICD-9 = 155 included HCC, cholangiocarcinoma and other malignant neoplasm of intrahepatic bile ducts. Only patients with a HCC diagnosis were enrolled in the study. Patients with a diagnosis other than HCC (such as cholangiocarcinoma) were not included.
Each patient's name, chart number, national citizen identification number, birthday, sex, residency, zip code, the diagnostic criteria for HCC, definitive diagnosis date, HBsAg status and anti-HCV status were recorded. Since patients might have undertaken multiple visits to different hospitals, we used their national citizen identification numbers to delete any duplicated records.
Diagnosis of HCC
We arbitrarily classified the diagnostic criteria of HCC as 1–4. Criterion 1 applied to the diagnosis of HCC that was verified by either pathology or cytology. Criterion 2 involved the diagnosis of HCC that was based on an α-fetoprotein level higher than 400 ng/ml plus at least one imaging study that showed a typical HCC image. Criterion 3 was used for the diagnosis of HCC that did not fit criteria 1 or 2 initially but fit either criteria 1 or 2 during the follow-up period. Criterion 4 was used when the diagnosis of HCC was based on typical image studies but did not fit criteria 1–3. HCC patients with diagnostic criteria 1–3 were grouped in the definite diagnosis group, while HCC patients with diagnostic criterion 4 comprised the clinical diagnosis group.
Etiology of HCC
B-HCC was defined as HCC with a positive HBsAg status and a negative anti-HCV status. C-HCC was defined as HCC with a negative HBsAg status and a positive anti-HCV status. The HBV plus HCV-related HCC (B+C-HCC) was defined as HCC with a positive status for both HBsAg and anti-HCV. The non-HBV and non-HCV-related HCC (NBNC-HCC) was defined as HCC with a negative status for both HBsAg and anti-HCV.
Figure 2 illustrates the flow chart for the analysis of secular trends and geographic variations. Anti-HCV status was not available for most cases diagnosed before 1994; if stored sera were available from the sera bank, anti-HCV status was tested retrospectively. Anti-HCV data could not be obtained for a number of cases diagnosed before 1994. Thus, HBsAg was used as marker for the analysis of secular trends. We classified HCC into 2groups, B-HCC and non-B-HCC. The annual B-HCCs mortalities were estimated from the HBsAg positive rate times the official age-adjusted death rate.
The geographic variation was analyzed in 23 counties and cities. In our previous studies of HCC patients, we found that the geographic variation in HBsAg prevalence was smaller than that in anti-HCV prevalence.18 Thus, we focused on the latter. We included patients from 1994 to 2001 for analysis because anti-HCV data were not universally available in Taiwan before 1994. Anti-HCV prevalence and 95% confidence intervals were calculated. We used goodness of fit to compare the anti-HCV prevalence in counties and cities with the mean anti-HCV prevalence for Taiwan as a whole. City- and county-specific prevalence of anti-HCV were presented as median, ranges and standardized mortality ratios (SMR, defined as anti-HCV prevalence of HCC patients of a certain city or county divided by anti-HCV prevalence of HCC patients in Taiwan as a whole).
Foxpro® was used to create the database and SPSS® (10.0) was used for statistical analysis. We used the Student's t test, one-way ANOVA, linear correlation, goodness of fit and χ2 test for a linear trend to evaluate the mean, standard deviations for continuous variables, percentages and 95% confidence intervals for nominal variables. The statistical significance was defined as p < 0.05.
Demography of patients
A total of 18,423 cases were reviewed from the 8 hospitals. Among them, 6,743 (36.6%) fitted diagnostic criteria 1; 5,399 (29.3%) fitted diagnostic criteria 2 and 648 (3.5%) fitted diagnostic criteria 3. Thus, 12,790 (69.4%) was classified as the definite diagnosis group, while 5,633 (30.6%) was classified as the clinical diagnosis group. HBsAg was available for 12,616 of these 12,790 cases, made up of 10,149 males and 2,467 females (male/female ratio: 4.1). The mean age for male HCC patients was 56.3 ± 13.5 years, while it was 60.3 ± 14.0 years for female.
To compare the differences in the gender, age and etiology between the definite diagnosis group and the clinical diagnosis group, we analyzed the patients with available HBsAg and anti-HCV data (years 1994–2001). These patients included 8,595 patients (6,741 male and 1,854 female) in the definite diagnosis group, and 3,274 patients (2,513 male and 761 female) in the clinical diagnosis group (Table I). The patients in the definite diagnosis group with a mean age of 57.9 years were on an average younger than patients in the clinical diagnosis group (57.9 ± 13.5 vs. 60.6± 12.2, p < 0.001).
Table I. Differences in Age, Sex and Viral Etiology Between the Definite Diagnosis Group and the Clinical Diagnosis Group
There was a borderline statistically significant difference in the sex ratio between these 2 groups (p = 0.049). Regarding the etiology of the HCC, more NBNC-HCC was found in the clinical diagnosis group than in the definite diagnosis group (16.8% vs. 10.7%, p < 0.001).
Etiology, gender and age in HCCs
To investigate the distribution of gender, age and etiology for the HCC patients, we analyzed the patients in the definite diagnosis group with available HBsAg and anti-HCV data (years 1994–2001). As shown in Table II, there were statistically significant differences in the mean age and sex ratio between the B-HCC and C-HCC groups. The mean age of B-HCC patients was 12 years younger than that of C-HCC patients (53.2 ± 13.6 years vs. 65.1± 9.1 years). The male/female ratio in B-HCC was 6.4, while it was 1.7 in the C-HCC (p < 0.001). The predominant etiology of male HCC was HBV. In contrast, the predominant cause of female HCC was HCV. The percentages of B+C-HCC and NBNC-HCC were similar when male and female HCC patients were compared.
Table II. Differences in Age and Sex According to Etiology of HCC in the Definite Diagnosis Group1
Age (years) (mean ± SD)
B, HBV-related HCC; C, HCV-related HCC; B+C, HBV plus HCV-related HCC; NBNC, non-HBV and non-HCV-related HCC.
Patients with available HBsAg and anti-HCV data (years 1994 to 2001) were only included.
B (N = 4,569)
53.2 ± 13.6
52.9 ± 13.3
55.0 ± 14.7
3949/620 = 6.4
B+C (N = 713)
59.0 ± 11.6
57.8 ± 11.9
63.6 ± 9.4
564/149 = 3.7
C (N = 2,396)
65.1 ± 9.1
64.6 ± 9.4
66.0 ± 8.5
1520/876 = 1.7
NBNC (N = 917)
62.1 ± 14.1
62.3 ± 13.6
61.3 ± 15.7
708/209 = 3.4
All (N = 8,595)
57.9 ± 13.5
56.9 ± 13.5
61.6 ± 12.9
6741/1854 = 3.6
Secular trends for HBV-related HCC
The percentage of B-HCC in male HCC patients progressively decreased from 1981 to 2001. The trend was more evident if we used 3-year intervals and tested by χ2 test for a linear trend (Fig. 3). The average percentage of B-HCC had decreased from 81.5% in 1981–1983 to 66.2% in 1999–2001. To investigate whether the decline in the percentage of B-HCC was caused by the decrease in an absolute amount of B-HCC, we used aged-adjusted HCC death times the percentage of B-HCC in the corresponding year. As shown in Figure 4, the total male HCC death rate progressively increased from 1981 to 2001. However, the B-HCC death rate remained stationary. This meant that there was a progressive decrease in the percentage of male B-HCC, caused by an increase in non-B-HCC. Indeed, the death rate from non-B-HCC increased 2-fold, from 6.2 × 105 in 1981–1983 to 13.4 × 105 in 1999–2001. When the anti-HCV test became universally available after 1993, it was found that most of the increase in the non-B-HCC mortality was related to HCV infection.
In a similar way to their male counterparts, the percentage of B-HCC in female HCC patients progressively decreased from 1981 to 2001. The trend was more evident if we used 3-year intervals and tested by χ2 test for a linear trend (Fig. 5). The average percentage of B-HCC progressively decreased from 66.7% in 1981–1983 to 41.4% in 1999–2001. Total female HCC death also progressively increased from 1981 to 2001, and death from non-B-HCC also increased from 3.39 × 105 in 1981–1983 to 8.04 × 105 in 1999–2001 (Fig. 6). However, in contrast to male HCC, there was a slight decline in B-HCC deaths among female HCC patients (Fig. 6). This meant that the progressively decreasing percentage of female B-HCC was caused by the increase in the non-B-HCC plus the slight decrease in B-HCC. When the anti-HCV test became was universally available after1993, we were able to establish that most of the increase in the non-B-HCC mortality was related to HCV infection.
Secular trends in ages of HCC patients
The mean diagnostic ages of HCC patients progressively increased over the past 20 years. The mean age was 50.0 ± 14.3 years in 1981 and increased to 58.6 ± 13.5 years in 2001 (Fig. 7). Comparisons of the mean age of HCC patients for 1981–1990 withthe mean age for 1991–2001 showed that patients were statistically significantly younger over the period 1981–1990 for either total HCCs (54.4 ± 14.0 years vs. 57.9 ± 13.5 years), B-HCCs (52.1 ± 13.7 years vs. 54.0 ± 13.5 years) or non-B-HCCs (61.6 ± 12.4 years vs. 64.2 ± 10.9 years) (p < 0.001). The same trend also held true if we restricted the cases to male or female HCC patients.
Geographic variations in anti-HCV prevalence of HCC patients
The mean anti-HCV prevalence for male HCC patients in Taiwan was 31.5% (95% confidence interval: 30.4∼32.6%). Anti-HCV prevalence for male HCC patients was statistically significantly higher in Chiayi city (52.9%, SMR = 1.68), Chiayi county (51.4%, SMR = 1.63), Yunlin county (44.5%, SMR = 1.41), Tainan county (41.8%, SMR = 1.33) and Kaohsiung county (35.9%, SMR = 1.14) (Fig. 8a). Anti-HCV prevalence was statistically significantly lower in Hualien county (18.6%, SMR = 0.59), Taoyuan county (18.7%, SMR = 0.59), Taitung county (19.1%, SMR = 0.6), Taipei county (19.1%, SMR = 0.61), Pingtung county (23.5%, SMR = 0.75) and Taipei city (25.1%, SMR = 0.8).
The mean anti-HCV prevalence for female HCC patients in Taiwan was 56.7% (95% confidence interval: 54.4∼59.0%). Anti-HCV prevalence for female HCC patients was statistically significantly higher in Chiayi city (87.8%, SMR = 1.55), Yunlin county (68.3%, SMR = 1.2) and Tainan county (67.2%, SMR = 1.18) (Fig. 8b). Anti-HCV prevalence was statistically significantly lower in Pingtung county (38.1%, SMR = 0.67) and Taipei city (46.4%, SMR = 0.82).
One of the most important findings in our study was that the percentage of B-HCC patients progressively decreased over the past 20 years. The relative decrease in B-HCC was not due to decreased B-HCC death. Instead, it was the result of an increase in non-B-HCC patients. As shown here, most non-B-HCCs in Taiwan after 1993 were related to HCV infection and, therefore, the relative decrease B-HCC was most likely caused by an increase in C-HCC. This trend is similar to trends described in countries with low HBV infection rates, such as Japan20, 21, 22 and the United States.23 Available studies indicate that HCV infections acquired 2–4 decades ago explain at least half of the observed increase in HCC in Western countries, and C-HCC is quite likely to continue to increase over the next decade.24, 25 Our results exhibited that an increase in C-HCCs in HBV-endemic area like Taiwan. Thus, in Taiwan and possibly in other areas of the world, prevention of HCV infection and treatment of chronic hepatitis C should become major goals of HCC prevention programs.
Several potential sources for bias exist in our study. The study contains hospital-based cases only, which may lead to potential bias. However, since there was no established patient-referral mechanisms in the medical system in Taiwan, patients were able to visit any hospital at any time at their will. Thus, the case mix of these hospitals is quite likely to represent the target populations in Taiwan. Increased awareness of the anti-HCV testing might have increased the proportion of C-HCC. However, if we stratified the HCC patients into 2 groups, B-HCC and non-B-HCC, the case number and proportion of non-B-HCC increased. As the majority of non-B-HCC was C-HCC, this implied that increased awareness of anti-HCV testing was not a major factor leading to an increase in C-HCC patients. The proportion of missing anti-HCV data after 1994 in our patients was 4.8%. Thus, missing anti-HCV data were probably not a major confounding factor in the secular trends. The HCC patients of clinical diagnosis group were not included in the analyses of secular trends and geographic variations. Exclusion of such patients might lead to a potential bias. The percentage of NBNC-HCC in the clinical diagnosis group of our study was 16.5%, which was higher than that in the definite diagnosis group. It is possible that some of these cases in the clinical diagnosis group were not true HCC.
Our study found geographic variations in anti-HCV prevalence of HCC patients in the small island of Taiwan. Previous smaller scaled studies have identified several high-HCV infection areas in Taiwan.18 The high HCV infection rate in these areas might be related to the iatrogenic routes of infection. Local customs and medical treatment seeking-behaviors, such as frequent intravenous injections for minor disease as well as incomplete disinfection of medical equipments in the past, might have contributed to the high HCV prevalence.19, 26
However, we should point out that HCV prevalence is not always correlated with the HCC prevalence. In a previous study, we found two HCV-endemic townships, Paisha and Tzukuan. All HCC cases from Paisha were HBsAg positive, while 13/14 HCC cases from Tzukuan were anti-HCV positive. The endemic nature of HCV in Tzukuan appears to have been long enough to induce HCC, but HCV turns out to be a newly introduced infection in Paisha.19 Thus, the secular trends and geographic variations of B-HCC and C-HCC do not represent the current state of infection, but the past state of infection. Since this study enrolled only HCC patients, we may have missed some endemic areas where HCV is a newly introduced infection. In this regard, active screening methods to identify HCV-infected endemic areas are important in the future prevention of HCC.
Differences existed in the mean age and gender ratio among HCCs of different viral etiologies. The mean age of B-HCC patients was 12 years younger than that of C-HCC patients, which is similar to previous results.18, 21 This difference may be explained by the natural course of chronic HBV or HCV infection; that is, HBV infection is usually acquired perinatally, while HCV infection is acquired in adulthood. The overall male/female ratio in our study was 3.6. However, the male/female ratio of B-HCC patients was 6.4, which is much higher than that the 1.7 male/female ratio of C-HCC patients. Sex hormone levels have been implicated in relation to hepatocarcinogenesis,27, 28 this might partially explain why B-HCC occurs predominantly in men. Because the majority of B-HCC was male, there were relative few female B-HCC patients. Therefore, the most common etiology in female HCC patients turned out to be HCV. The mean age and sex ratio for B+C-HCC were positioned between those for B-HCC and C-HCC. The exact explanation for this observation is still unknown. However, we hypothesize that B+C-HCC patients, like B-HCC patients, are infected with HBV early in life. The HBV becomes inactive as time passes, leading to less liver damage and possible less risk of HCC. Thus, the mean age when HCC developed is older than the age at which B-HCC developed. The B+C-HCC patients subsequently acquire HCV infection later in life. HCV takes over from HBV and becomes the major player leading to hepatitis activity.29, 30 Since the prior HBV infection has already induced some hepatitis and/or fibrosis in the liver, the superimposed HCV infection leads to further and thus faster damage to the liver from HCV. Thus, the mean age of B+C-HCC is younger than that of C-HCC. Our hypothesis needs to be explored further in the future.
Although we defined HBsAg-negative, anti-HCV-negative HCC as NBNC-HCC, earlier studies have showed that some of these NBNC-HCC are caused by an occult HBV infection.31, 32 The presence of anti-HBc is useful in diagnosing HBV-related HCC. However, anti-HBc data were not available for most of our NBNC-HCC cases. The mean age of the NBNC-HCC group was closer to that of the C-HCC group, while the sex ratio of the NBNC-HCC group was closer to that of the B+C-HCC group. Thus, whether most of the NBNC-HCC was really related to HBV infection needs to be further clarified. Alcoholism may be one of the etiologies of the NBNC-HCC.33 However, because this was a retrospective study, it was difficult to verify alcoholism among the NBNC-HCC patients.34, 35
In conclusion, our study demonstrated that in Taiwan approximately 67% of HCC in men was related to HBV infection whereas about 55% of HCC in women was related to HCV infection. The mean age for B-HCC was 12 years younger than for C-HCC. The mean age of male HCC patients was 5 years younger than that of female HCC patients. The percentage of B-HCC has progressively decreased over the last 20 years. However, the relative decrease in B-HCC was not due to decreased B-HCC death. Instead, it was the result of an increase in C-HCC. C-HCC was more prevalent in central and southern Taiwan than in northern Taiwan. Prevention of new HCV infection and treatment of chronic hepatitis C should be major goals for HCC prevention in the future, which is attainable even in an HBV-endemic area such as Taiwan.
We are indebted to our colleagues at the Department of Medical Records for their excellent work in the cancer registry system and the physicians for their care of the patients.