Tel.: +82-2-740-8338, Fax: 82-2-747-4830
Coinfection of hepatitis B and C viruses and risk of hepatocellular carcinoma: Systematic review and meta-analysis
Article first published online: 27 OCT 2010
Copyright © 2010 UICC
International Journal of Cancer
Volume 128, Issue 1, pages 176–184, 1 January 2011
How to Cite
Cho, L. Y., Yang, J. J., Ko, K.-P., Park, B., Shin, A., Lim, M. K., Oh, J.-K., Park, S., Kim, Y. J., Shin, H.-R., Yoo, K.-Y. and Park, S. K. (2011), Coinfection of hepatitis B and C viruses and risk of hepatocellular carcinoma: Systematic review and meta-analysis. Int. J. Cancer, 128: 176–184. doi: 10.1002/ijc.25321
- Issue published online: 27 OCT 2010
- Article first published online: 27 OCT 2010
- Manuscript Accepted: 25 FEB 2010
- Manuscript Received: 5 OCT 2009
- National Cancer Center Foundation. Grant Number: 0790860-1
- BRL (Basic Research Laboratory) program through the National Research Foundation of Korea funded by the Ministry of Education, Science and Technology. Grant Number: 2009-0087452
- hepatocellular carcinoma;
- sub-additive effect;
- hepatitis B virus;
- hepatitis C virus;
A subadditive effect of hepatitis B virus (HBV) and hepatitis C virus (HCV) coinfection is possible because superinfection of one virus tends to inhibit infection of the other virus. However, studies have reported inconsistent findings, and two meta-analyses of studies from various countries (1998) and China (2005) reported a supraadditive effect for hepatocellular carcinoma (HCC) risk. Thus, we reevaluate HBV/HCV monoinfection and coinfection. Of 411 reports, we included 59 studies that assessed the association between HBV/HCV monoinfection and coinfection for HCC risk. HCC risk because of high/detectable HBV DNA and HBeAg infection was higher than HBsAg infection, whereas anti-HCV vs anti-HCV/HCV RNA was not different. Geographically, HCC risk was significantly higher in nonendemic than in HBV or HCV endemic areas. Subadditive effect for HCC risk was presented in recently published studies, cohort studies and studies conducted in HBV/HCV nonendemic areas; an additive effect was presented in studies conducted in HBV endemic areas; a supraadditive effect was presented in previously published studies, case-control studies and studies conducted in HCV endemic areas. Our results suggest HBV/HCV coinfection for HCC risk is not significantly greater than HBV/HCV monoinfection, and HCC risk due to HBV or HCV is higher in nonendemic than endemic areas. The p-heterogeneity was significant for most analyses, except HBV(+)/HCV(+) and HBV biomarker analyses. Prevention strategies targeted toward HBV or HCV monoinfected patients are needed. In addition, tailored prevention to reduce infectivity such as HBV markers (HBeAg, HBV DNA) is needed.
Hepatocellular carcinoma (HCC) is the sixth most common cancer and the third most common cause of mortality from cancer worldwide.1 Hepatitis B virus (HBV) and hepatitis C virus (HCV) monoinfections are well-known major risk factors for HCC, and the relative importance varies worldwide and changes over time.2 Because of their shared modes of transmission, coinfection of HBV and HCV is not uncommon, particularly in countries with a high prevalence of HBV or HCV. HBV and HCV coinfection results in more severe liver disease3 and in an increased risk of HCC4 than monoinfection.
Epidemiologic studies on viral interaction have not been consistent. Some reported no interaction5–10; others reported a sub/supraadditive or multiplicative interaction.11–18 Two meta-analyses, one based on case-control studies from various countries published before 1998 (32 studies with 4,560 cases vs 6,998 controls)19 and the other based on case-control studies from China published before 2005 (32 studies with 3,201 cases vs 4,005 controls),20 reported supraadditive effects. However, because superinfection of one virus tends to inhibit infection of the other virus among coinfected cases,21, 22 a subadditive, rather than a supraadditive, effect is reasonable. The reports of a supraadditive effect are likely because of the following: (i) selection bias because of inclusion of only case-control studies; (ii) misclassification bias because of inclusion of only HBsAg and predominantly anti-HCV biomarkers that do not consider infectivity and viral molecules (HBeAg, HBV DNA and HCV RNA); and (iii) small number of controls with coinfection. In recent years, large population cohort studies on HBV and HCV monoinfection and coinfection15, 17, 23–28 and those using biomarkers such as HBeAg and HBV DNA related to HBV infectivity and HCV RNA16 have been published.
Moreover, nearly 20 yr have passed since the introduction of the hepatitis B vaccine, and those countries that have integrated the hepatitis B vaccine into their national immunization programs29 have seen a decline in the prevalence of HBsAg. In addition, HCC risk by HBV infection has also decreased. This study also evaluated whether HCC risk by HBV infection has declined more in recent years (2000–2009) relative to previous years (1991–1999).
The objectives of this meta-analysis were to include recently published studies to reevaluate HCC risk due to HBV/HCV monoinfection and coinfection and to conduct subanalyses according to various categories, such as time period, biomarkers of current infection, study design and geographical region. As previously noted, because superinfection of one virus tends to inhibit infection of the other virus among coinfected cases,21, 22 a subadditive, rather than a supraadditive, effect is reasonable. We also assessed whether HCC risk by HBV infection has declined more in recent years (2000–2009) relative to previous years (1991–1999).
Material and Methods
Literature search and criteria for the selection of paper
The data retrieved for the meta-analysis were based on published case-control and cohort studies that examined the effect of HBV and HCV monoinfection and coinfection on the development of HCC. Figure 1 presents a schematic diagram of the literature search process, including inclusion and exclusion criteria. The Pubmed, Embase and Medline databases for articles (at least the abstract) published in English from January 1985 to June 30, 2009, were searched using keywords, i.e., hepatocellular carcinoma, liver cancer, hepatitis B virus and hepatitis C virus. A full manual search from references of selected articles was performed for further relevant publications.
Inclusion criteria for the systematic review were as follows: (i) case-control studies that recruited HCC cases and controls without chronic liver disease; (ii) cohort studies conducted among healthy individuals; (iii) studies that calculated the odds ratio (OR)/relative risk (RR) or provided raw data that allowed for the computation of the crude OR/RR estimates; and (iv) Studies that used HBsAg/HBeAg/HBV DNA and anti-HCV/HCV RNA serological markers. When a study contained data on more than one serological marker, HBsAg and anti-HCV were used in the main analysis.
Studies that classified subjects according to IgG anti-HBc or anti-HBs antibodies were excluded because these are markers of past infection,30 which was not the interest of the current study. In addition, studies that examined HCC cases without a comparison group or those that selected participants according to HBsAg/HBeAg/HBV DNA or anti-HCV/HCV RNA seropositivity were excluded. When more than one study from the same study population was available, the study with the most comprehensive data was used. Information for all serological biomarkers was included in the biomarker subanalysis.
Mantel-Haenszel crude estimates of the OR/RR and corresponding 95% confidence interval (CI) were calculated when the OR/RR was not presented but sufficient data were available. Logit OR/RR estimates were calculated when the data were sparse. If the article reported stratified OR/RR estimates, the strata were combined and the crude OR/RR was recalculated. Calculation of the OR/RR estimates was performed using SAS software version 9.1 (SAS Institute, Cary, NC). Random-effect model was used to obtain the summary OR and 95% CI. Heterogeneity was assessed by the Cochran Q statistics,31 and publication bias was assessed according to the Egger regression asymmetry test and the Begg and Mazumdar adjusted rank correlation tests.32, 33
To determine the source of heterogeneity, subgroup analyses, according to the following characteristics, were performed: (i) time period; (ii) biomarker; (iii) study design; and (iv) geographical area. Studies published before 2000 (1991–1999) or since 2000 (2000–2009), case-control and cohort studies designs were compared. With respect to biomarkers, studies that used HBsAg vs HBsAg and/or HBV DNA and HBeAg, and anti-HCV vs anti-HCV and/or HCV RNA were compared. In terms of geographical area, studies were categorized according to endemicity, similar to previous meta-analyses.19, 20 Geographical Area 1 was HBV endemic countries (i.e., China, Taiwan, Korea, Gambia, Thailand, Vietnam, Saudi Arabia, Morocco, Sudan and South Africa); geographical Area 2 was countries where HCV is predominant among HCC cases (i.e., Japan, Greece, Italy, Spain and Egypt); geographical Area 3 was countries with low HBV and HCV prevalence rates and minor role of both infections (i.e., United States and Australia).
Subadditive (less than the sum of the OR point estimates for HBV and HCV monoinfections), additive (equal to the sum), supraadditive (greater than the sum but less than the product) and multiplicative (equal to the product) interactions were determined by examining the point OR estimate. We additionally took into account the 95% CI and, thus, (i) calculated the summary OR (95% CI) for HBV and HCV monoinfection and (ii) determined the p value for heterogeneity of the summary OR (95% CI) in (i) and the OR (95% CI) for HBV/HCV coinfection according to Cochran Q statistics.31 Supraadditive, additive and subadditive effects were defined by the Cochran Q statistic p value and by the difference in coinfection OR estimate and sum of the OR point estimate for monoinfection. Supraadditive effect was defined when the pcochran <0.1 and the OR point estimate for coinfection was greater than the sum of the OR point estimates for monoinfection. Additive effect was defined as pcochran <0.1, and the OR point estimate for coinfection was similar to the sum of the OR point estimates for monoinfection. Subadditive effect was defined when the pcochran >0.1, and the OR point estimate for coinfection was less than the sum of the OR point estimates for monoinfection. All meta-analyses were conducted using STATA (version 10; Stata, College Station, TX).
Table 1 demonstrates the risk estimates for HBV and HCV mono-infection (Supporting Information Appendix Ia–Ic and IIa–IIc) and subanalyses according to time period, biomarker, study design and geographical area. Forty-seven studies (37 case-control and 10 cohort studies) examined the association between HBsAg/HBV DNA and risk of HCC. The majority of these studies showed a significantly positive association for HCC risk with an OR that ranged from 3- to 75-fold for HBsAg; three studies34–36 showed a marginally significant association. For HCV infection, 41 studies (37 case-control and 4 cohort studies) assessed the association between anti-HCV and/or HCV RNA and HCC risk. The majority of the studies showed a significantly positive association (OR = 4–241). Two studies reported marginal significance6, 13; two studies reported insignificant association37, 38; and three studies12, 39, 40 reported insignificant associations for anti-HCV but significant association when HCV RNA was added as a biomarker.12, 39, 40P-heterogeneity was significant (p = 0.001) for subanalyses of HCV time trend effect, HBV biomarker and HCV geographical area. HCV-related HCC risk was marginally lower for studies conducted in 1991–1999 compared to 2000–2009 (pheterogeneity = 0.056). [HBsAg and (HBV DNA or HBeAg)] was significantly higher than HBsAg only. For geographical area subanalysis of HCV studies, the OR (95% CI) for Areas 1, 2 and 3 were 8.3 (5.4–12.7), 13.8 (7.8–24.5) and 23.8 (16.9–33.5), respectively. There was publication bias (Egger, p ≤ 0.05) in three of the subanalyses; Figures 2a–2c present the Begg funnel plots for HBV studies published 1985–1999 (Fig. 2a), all HCV studies (Fig. 2b) and anti-HCV and/or HCV RNA (Fig. 2c). There was heterogeneity across the studies (Cochran Q test, p < 0.05) before and after stratification, except studies using HBsAg and [HBV DNA or HBeAg].
Table 2 summarizes the risk estimates of HBV and HCV coinfection for HCC risk (Supporting Information Appendix IIIa and IIIb) stratified by time period, study design and geographical area. Twenty-two studies (20 case-control and 2 cohort studies) assessed the relationship between HBsAg and/or HBV DNA and anti-HCV/HCV RNA coinfection and risk of HCC. Studies reported subadditive,5, 8–10, 13, 17, 18, 24, 34, 41–46 supraadditive11, 12, 14, 16, 47, 48 and multiplicative risk interactions.49 In our meta-analysis of all studies, HBV/HCV coinfection presented an additive effect for HCC risk. According to time period, HCC risk for more recently published studies showed a subadditive effect, whereas previously published studies showed a supraadditive effect. Coinfection in case-control studies had a supraadditive effect, whereas cohort studies had a subadditive effect. For geographical area, HBV endemic areas (Area 1) presented an additive effect; HCV endemic areas (Area 1) presented a supraadditive effect. A meta-analysis for nonendemic areas (Area 3) was not performed because of the small number of studies (n = 1), which showed a subadditive effect. There was publication bias (Egger, p ≤ 0.05) for HBV(+)/HCV(−) for studies published in 2000–2009 (Fig. 2d) and heterogeneity across the studies (Cochran Q test, p < 0.05) before and after stratification in much of the analysis.
Supporting Information Appendix IV summarizes the cofactors presented in the studies. For some of the studies, there were differences in age, gender, alcohol and smoking history between the comparison groups (p < 0.05 as indicated with an asterisk, see footnote). Three studies included information on HCV genotypes and showed genotypes varied among countries.
Our main findings demonstrated HBV and HCV infections are independent risk factors with a 12- to 14-fold increased risk for HCC. High/detectable HBV DNA and HBeAg increased HCC risk up to 31-fold. HBV-related HCC risk did not show a significant time trend effect, whereas HCV-related HCC risk showed a marginally significant increased time trend effect. Geographically, among HCV studies, HCC risk was significantly higher in nonendemic than HBV or HCV endemic areas. Subadditive effect for HCC risk was presented in recently published studies, cohort studies and studies conducted in HBV/HCV nonendemic areas; an additive effect was demonstrated in studies conducted in HBV endemic areas; and a supraadditive effect was shown in previously published studies, case-control studies and studies conducted in HCV endemic areas.
Our results showed that biomarkers, such as HBeAg and HBV DNA, which assess viral infectivity, had a greatly significantly increased risk for HCC. Subjects with both HBsAg and HBeAg, biomarkers for active HBV replication and transmission,16, 50, 51 had a highly increased risk of HCC.16 Elevation in serum HBV DNA, a sensitive biomarker that shows liver disease activity and viral replication,50 was associated with an increased risk of HCC, independent of HBeAg.52 An Italian study reported that anti-HCV(+)/HCV RNA(+) subjects had an increased risk compared to anti-HCV(+) subjects.44 In our meta-analysis, there was no heterogeneity between the summary OR (95% CI) for anti-HCV and [anti-HCV and HCV RNA], whereas there was heterogeneity for HBsAg vs. HBsAg and [HBV DNA or HBeAg]. Therefore, tailored prevention strategies to primarily reduce viral replication detected with HBeAg and HBV DNA are needed.
In 1992, the World Health Organization recommended countries with HBV carrier prevalence greater than 8% integrate the hepatitis B vaccine into their national immunization programs within 3 yr, by 1995.53 As the HBV vaccine reduced HBV prevalence in Taiwan,54, 55 Gambia56 and Malaysia,57 a reduction in HCC risk was consequently expected. However, although HCC risk from the 2000–2009 studies was slightly lower than the 1991–1999 studies, this was not a significant reduction in our analysis.
On the contrary, a marginally significant (pheterogeneity = 0.056) trend effect of HCV-related HCC risk was demonstrated, which may partly reflect improvements in detection or reclassification because of increased awareness of HCC risk, more specific diagnostic criteria and greater sensitivity of anti-HCV enzyme-linked immunosorbent assay tests. In the United States, the HCV epidemic began in the 1960s and peaked in the 1980s58; the incidence of HCC began increasing in the mid-1980s, accelerating in the 1990s and continuing to rise.59 Although HCV incidence did not culminate at exactly the same time in all parts of the world, a peak incidence did occur in the 1980s, rather than the 1970s, which may partly explain the increase in HCC risk.59 In addition, since the 1960s, there has been an increase in HCV due to intravenous drug abuse, which is expected to increase the incidence of HCV-related HCC.60
Our results showed that geographically HCC risk in nonendemic areas was higher than in HBV or HCV endemic areas, which may partly be due to mostly case-control studies where the OR is calculated as [odds of exposure in cases/odds of exposure in controls]. In endemic areas, the denominator and numerator are concurrently increased because exposure is high in both cases and controls. However, in nonendemic areas, HCC cases exposed to HBV or HCV infection is high because of HBV or HCV carcinogenic activity, whereas HBV or HCV exposed controls is low. Therefore, the ratio of the numerator to the denominator increases. Other explanations are possible. A greater number of cases due to environmental factors other than hepatitis viruses such as Clonorchis sinensis,61–63 alcohol12, 64, 65 and smoking may have diluted the summary OR in endemic areas. In addition, age and transmission mode affect the severity and chronicity of hepatitis. In endemic regions, such as Asia and Africa, HBV infection is often acquired early in life either via perinatal transmission or contact with other infected children.66 In contrast, in nonendemic areas, such as Northwestern Europe, North America and Australia, HBV infection occurs mainly by sexual contact and IV drug use with a peak incidence between the ages of 15 and 25 yr.67 And, older aged subjects can progress more easily to chronic hepatitis.68
The supraadditive effect reported in the earlier years (1991–1999) may be due to several studies without coinfection controls (n = 0; Supporting Information Appendix IIIb) that shifted the summary OR (95% CI) for HBV/HCV coinfection to the right. In addition, this may have resulted in inconsistency in interaction effects in the subanalyses. Considering current biological knowledge, a subadditive effect of HBV and HCV coinfection is reasonable. Clinical and experimental studies demonstrated a reciprocal interference of one virus on the replication of the other.21, 22, 69, 70 Clinical studies showed coinfected cases had a HBV dominant or HCV dominant effect.4, 71 In chronic dual or triple HCV, HBV (+/−HDV) coinfection, viruses displayed reciprocal interference on viral genome replication, and viral DNA or RNA was undetectable in serum.72, 73In vitro studies reported inhibition of HBV replication may be HCV genotype dependent and mediated by HCV core protein.52, 74 Chimpanzees with chronic HBV and acute HCV superinfection had reduced serum HBsAg titer.69, 70 HCC prevention strategies targeted toward patients with HBV or HCV monoinfection are needed in endemic areas. In nonendemic areas, HCC risk because of HBV/HCV coinfection may be greater than expected, and thus prevention strategies primarily targeted toward preventing blockage of the transmission pathway and initial elimination of HBV or HCV are needed.
The summarized cofactors presented in the studies (Supporting Information Appendix IV) showed that there were differences in age, gender, alcohol and smoking history between comparison groups in some studies that may affect the summary OR (95% CI) for the meta-analysis. However, when available, our study used the adjusted OR (95% CI; Supporting Information Appendix I–III), thereby minimizing the confounding affect that may result otherwise.
Limitations of the study are noted. First, most studies used HBsAg as a biomarker for HBV infection, except a few studies that used HBsAg and/or HBV DNA9, 75 or HBeAg.14, 16, 48, 65, 76 Reports described subjects with chronic HCV infection who had occult or clinically silent HBV infection (undetectable HBsAg with detectable HBV DNA in serum) detected by highly sensitive molecular techniques.73, 77 Thus, HBsAg as the only HBV biomarker may substantially underestimate the number of HBV/HCV coinfected subjects. Second, most subanalyses had statistically significant heterogeneity except HBV biomarkers and HBV(+)/HCV(+) analyses. Therefore, our results should be interpreted with caution. Selection bias partly contributed to the publication bias. Fortunately, in the HBV/HCV coinfection analyses, there was no publication bias except the subanalysis for time period (2000–2009). Third, other factors may modify the risk of HCC. Risk factors and etiology vary among geographical regions. For instance, aflatoxin B1 is a common contaminant of grain, nuts and vegetables in Asia and parts of Africa and is a risk factor secondary to hepatitis.60, 78, 79 Gender, age at infection, concomitant HIV infection, alcohol consumption and cigarette smoking may modify risk of HCV-related HCC.78 Many of the studies included in the meta-analysis presented male gender, older age, alcohol consumption and cigarette smoking with an increased HCC risk. Furthermore, variation in viral genotype may influence worldwide risk of HCC.80 The three studies that included information on HCV genotype (Supporting Information Appendix IV) showed that genotype variation differed according to country and that certain genotypes are associated with a significantly greater increased risk for HCC.
There are a number of strengths of the study. First, subanalyses, according to various categories, were conducted, and the p-heterogeneity values were calculated. Second, despite the difficulty in assessing the interaction effect due to the rarity of HBV and HCV coinfection among subjects without HCC, this meta-analysis included 143 controls with coinfection. Previous meta-analyses included eight19 or 12.20
In conclusion, our study demonstrated that HBV and HCV are independent risk factors for HCC risk and, HBV infection with high infectivity markers, such as HBV DNA and HBeAg, is highly associated with HCC risk. Geographically, HCC risk due to HBV or HCV is higher in nonendemic than endemic areas. Moreover, our study suggests coinfection of HBV and HCV has a subadditive risk for HCC.
- 9SEN virus infection and the risk of hepatocellular carcinoma: a case-control study. Am J Gastroenterol 2003; 98: 2500–4., , , , .Direct Link:
- 10The role of diabetes in hepatocellular carcinoma: a case-control study among United States Veterans. Am J Gastroenterol 2001; 96: 2462–7., , .Direct Link:
- 12Case-control study on hepatitis C virus (HCV) as a risk factor for hepatocellular carcinoma: the role of HCV genotypes and the synergism with hepatitis B virus and alcohol. Brescia HCC Study. Int J Cancer 1999; 81: 695–9., , , , , , , , , .
- 15Prospective study on the risk of hepatocellular carcinoma among hepatitis C virus-positive blood donors focusing on demographic factors, alanine aminotransferase level at donation and interaction with hepatitis B virus. Int J Cancer 2004; 112: 1075–80., , , , .
- 30Immunization Action Coalition: vaccination information for healthcare professionals.
- 40Risk factors analysis for hepatocellular carcinoma in patients with and without cirrhosis: a case-control study of 213 hepatocellular carcinoma patients from India. J Gastroenterol Hepatol 2007; 22: 1104–11., , , , , , .
- 41Prevalence of hepatitis B and C viral markers in patients with hepatocellular carcinoma in Korea. J Korean Cancer Assoc 1991; 23: 723–7., , , .
- 45UDP-glucuronosyltransferase 1A7 genetic polymorphisms are associated with hepatocellular carcinoma risk and onset age. Am J Gastroenterol 2005; 100: 1758–63., , , , , .Direct Link:
- 49A study on the prevalence of HBsAg and anti-HCV in patients with hepatocellular carcinoma: comparative study with healthy blood donors. Korean J Clin Pathol 1998; 18: 459–63., , , , .
- 61Epidemiologic study of Clonorchis sinensis infestation in a rural area of Kyongsangnam-do South Korea]. J Prev Med Public Health 2005; 38: 425–30., , , , , , , . [
- 63Clonorchiasis in Taiwan. Southeast Asian J Trop Med Public Health 1991; 22( suppl): 184–185..
- 68Epidemiology and natural course of HCV infection. Korea: The Korean Association for the Study of the Liver, 2004..
- 73Virological profiles in patients with chronic hepatitis C and overt or occult HBV infection. Am J Gastroenterol 2002; 97: 1518–23., , , , , , , , , , .Direct Link:
Additional Supporting Information may be found in the online version of this article.
|IJC_25321_sm_SuppAppendix.doc||1280K||Supporting Information Appendix.|
Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.