Journal of Hepatology 50 (2009) 1142–1154 www.elsevier.com/locate/jhep
Hepatitis C virus genotype 1b as a risk factor for hepatocellular carcinoma development: A meta-analysisq Sara Raimondi1,*, Savino Bruno2, Mario U. Mondelli3, Patrick Maisonneuve1 1
Division of Epidemiology and Biostatistics, European Institute of Oncology, Via Ripamonti 435, 20141 Milan, Italy 2 Department of Medicine, Liver Unit, Azienda Ospedaliera Fatebenefratelli e Oftalmico, Milan, Italy 3 Research Laboratories, Department of Infectious Diseases, Fondazione IRCCS Policlinico San Matteo and University of Pavia, Pavia, Italy
Background/Aims: Hepatitis C virus (HCV) is a known risk factor for hepatocellular carcinoma (HCC), but whether the risk varies among patients infected with different HCV genotypes is still controversial. We performed a meta-analysis to clarify whether the genotype 1b is associated with a higher risk of HCC than other genotypes. Methods: We identified 57 relevant papers through a literature search to December 2007 but, since age could represent a major confounder, we focused the meta-analysis on the 21 studies presenting age-adjusted risk estimates for HCV genotype 1b vs. other genotypes. We used random-effects models with the DerSimonian–Laird method and assessed heterogeneity between studies and publication bias. Results: Patients infected with HCV genotype 1b have almost double the risk to develop HCC than those infected with other genotypes (Relative Risk (95% Confidence Intervals) = 1.78(1.36–2.32)). The pooled risk estimate was somewhat lower when we restricted the analysis to the eight studies conducted in patients with liver cirrhosis (1.60;1.07–2.39) or considering the 36 studies presenting only crude data (1.63;1.30–2.06). In seven studies excluding patients with liver cirrhosis, the RR (95% CI) increased to 2.46(1.69–3.59). Conclusions: This meta-analysis suggests that HCV genotype 1b plays an important role in HCC development, especially in patients with early stage liver disease. Ó 2009 European Association for the Study of the Liver. Published by Elsevier B.V. All rights reserved. Keywords: Epidemiology; Liver cancer; Virology
1. Introduction Approximately 20% of patients with HCV infection will develop cirrhosis after a mean period of 20 years, and 1–4% of them will eventually develop hepatocellular carcinoma (HCC) on average 30 years after infection
Received 10 November 2008; received in revised form 13 January 2009; accepted 26 January 2009; available online 20 March 2009 Associate Editor: J.M. Llovet q The authors who have taken part in this study declared that they do not have anything to disclose regarding funding or conflict of interest with respect to this manuscript. * Corresponding author. Tel.: +39 02 57489377; fax: +39 02 57489922. E-mail address:
[email protected] (S. Raimondi).
[1,2]. HCC represents the most serious complication of Chronic Liver Disease (CLD) and the most frequent cause of death in patients with compensated liver cirrhosis (LC) [3]. Besides HCV infection, other well known risk factors for HCC include infection with hepatitis B virus (HBV), high ethanol intake, older age, male gender, tobacco smoking, LC, advanced fibrosis stage, high histological activity score and elevated alanine aminotransferase serum levels [4–9]. In this setting, the impact of HCV genotypes has also been evaluated in several studies. Despite some of them reported that patients infected with HCV genotype 1b have a higher risk of developing HCC than those infected with other genotypes, [10–14] other studies did not confirm this result [15–18]. As a consequence, no
0168-8278/$36.00 Ó 2009 European Association for the Study of the Liver. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.jhep.2009.01.019
S. Raimondi et al. / Journal of Hepatology 50 (2009) 1142–1154
consensus has emerged, and the role of HCV genotype in both accelerating the progression of the disease and as a risk factor for HCC remains to be established. We therefore decided to perform a meta-analysis of published studies, which permitted to evaluate the role of HCV genotype 1b in HCC development with a powerful approach; this enabled us also to assess the sources of inconsistencies and variability in the estimates of the association between HCV genotype 1b and HCC, and to explain whether they may be ascribed to the composition and characteristics of the study population or the methodological features of the performed studies.
1143
heterogeneity rather than to chance. Between-study heterogeneity was explored through sub-group analyses and meta-regression [29]. We conducted sub-group analyses limited to studies reporting data on HCV genotype 1b (therefore excluding studies that did not differentiate genotype 1b from other HCV-1 subtypes), to studies comparing genotype 1b with type 2, and to studies conducted in subjects with CLD or LC. In order to explore the heterogeneity among studies, we included the following variables in meta-regression: publication year, study location, study design, underlying liver disease, inclusion and percentage of subjects undergoing IFN-a therapy, inclusion and percentage of subjects with HBV co-infection, inclusion and percentage of subjects with history of alcohol abuse, genotyping technique, genotype comparison. Publication bias was graphically represented by funnel plot and assessed by Egger’s test [30]. The statistical analysis was performed using STATA software, version 8.2.
2. Materials and methods 3. Results Studies suitable for this meta-analysis had to meet the following inclusion criteria: they had to provide either the frequency of HCV genotypes 1b and non-1b in HCC cases and controls or a measure of Relative Risk (RR), with 95% Confidence Intervals (CI), for the association between HCV genotype 1b and HCC development. Papers that reported association for genotype 1 instead of 1b were also included, as they were conducted in countries where 1b was the most common HCV-1 subtype [19–21]. These latter studies were however excluded in sensitivity analyses. Studies based on patients infected with specific HCV genotype(s) were excluded, as well as those on patients who underwent liver transplantation since genotype 1b could be over-represented in patients with decompensated liver disease requiring liver transplantation [20]. We performed a literature search to December 2007 on PubMed, ISI Web of Science (Science Citation Index Expanded) and Embase, using combinations of the keywords ‘‘hepatitis C virus”, ‘‘HCV”, ‘‘genotype”, ‘‘hepatocellular carcinoma”, ‘‘hepatocarcinoma”, and ‘‘liver cancer”. The search was limited to human studies with no language or time restrictions. In addition, we reviewed the references from all retrieved articles and relevant reviews [20–25] to identify additional studies. After initial screening of abstracts and references, 161 potentially relevant papers were identified and full-text was retrieved for detailed evaluation. Of the 67 papers which met the inclusion criteria, 10 were excluded because they overlapped with other selected papers with larger samples. Of the 57 studies eligible for the meta-analysis, 21 presented adjusted risk estimates for the association between HCV genotype 1b and HCC, while 36 reported only crude data on the frequency of HCV genotypes in cases and controls (Tables 1 and 2). For each study we recorded information on publication year, study location, study design, period of accrual, type of control, genotype comparison, inclusion and percentage of patients treated with interferon (IFN-a) therapy, with HBV co-infection, or with a history of alcohol abuse, HCC diagnosis, and genotyping methods.
2.1. Statistical analysis When available, we retrieved adjusted risk estimates presented in the original papers. If not, we retrieved the frequencies of HCV genotypes in cases and controls, and calculated the corresponding studyspecific crude Odds Ratio (OR), with 95% CI for HCC risk. We excluded mixed and undetermined genotypes in OR estimation, wherever possible. For studies with zero cells in the two by two table, we calculated OR by adding 0.5 to each of the four cells. We used random-effects models with the DerSimonian–Laird method [26] to evaluate summary estimate of the overall association of HCV genotype 1b with HCC. Study-specific estimates were weighted by inverse of variance. We ignored the distinction between various measures of RR (i.e. OR and RR) on the assumption that HCC is sufficiently rare [27]. Homogeneity among studies was tested by the Q statistic, with significance level set at 0.10, and by I2, [28] which represents the percentage of total variation across studies that is attributable to
We decided to focus the meta-analysis on the 21 studies that presented an age-adjusted estimate for the association between HCV genotype 1b and HCC, or in which cases and controls were matched at least by age. We made this choice since it was argued that the association between genotype 1b and HCC could be, at least partially, explained by a cohort effect: age representing therefore an important confounder. However, in order to check and validate the results based on adjusted risk estimates, we performed an additional analysis on studies providing only crude data: these results were briefly reported and commented. 3.1. Description of the studies The main characteristics of the 21 studies presenting age-adjusted risk estimates were described in Table 1. The articles were published between 1996 and 2007 and referred to a study period that covered 30 years, from 1976 to 2005. Eleven studies were conducted in Asia, nine in Europe, and one in Australia. Nine were cohort studies, while the remaining 12 were case-control studies. Studies providing frequency of HCV genotypes in prevalent cases, observed in a certain hospital during a specific time period were grouped with the ‘‘case-control” studies for the purpose of heterogeneity exploration. Five studies compared HCV genotype frequency in HCC cases and in HCV carriers without a specific CLD (asymptomatic subjects or patients with miscellaneous CLD); the remaining 16 papers focused only on patients with specific CLD: chronic hepatitis (CH) and/or LC. Among these 16 studies, eight articles included data on LC patients, with or without HCC. The laboratory techniques used to identify HCV genotypes varied among studies: PCR with genotype-specific primers was used in 11 studies, HCV Line-Probe-Assay (INNO-LiPA HCV) in six studies or both methods in two studies. In the remaining two papers HCV genotypes were identified by serological tests or by restriction fragment length polymorphisms analysis.
1144
S. Raimondi et al. / Journal of Hepatology 50 (2009) 1142–1154
Table 1 Description of the 21 studies presenting adjusted Relative Risk (RR) with 95% Confidence Interval (CI) for the association between HCV genotype 1b and HCC development. Author (year)
Country
Study design
N cases
N controls
Study population
RR (95% CI)a
Adjusting or matching variables
Hatzakis (1996) [57] Silini (1996) [10] Tanaka (1996) [12] Kuwana (1997) [55]
Greece Italy Japan Japan
Case-control Case-control Case-control Case-control
17 162 56 133
110 162 23 210
CH LC No liver disease CH and LC
8.3 1.7 3.8 2.6
Lopez-Labrador (1997) [63] Naoumov (1997) [59]
Spain
Case-control
50
50
LC
2.1 (0.4–12.0)
Age, gender Age, gender, child class Age, gender Age, gender, alcohol, smoking, blood transfusion Age, gender
UK
Case-control
30
17
LC
0.7 (0.2–2.3)
Niederau (1998) [64]b
Germany
Cohort
14
551
HCV liver disease
0.2 (0.1–1.1)
Tanaka H (1998) [32] Tanaka K (1998) [33]
Japan Japan
Case-control Cohort
179 25
972 24
No liver disease LC
2.3 (1.0–5.1) 0.6 (0.2–2.0)
Dutta (1999) [62] Murakami (1999) [56] Tagger (1999) [49]
Australia Japan Italy
Case-control Cohort Case-control
11 23 122
32 173 30
CH and LC CH and LC No liver disease
5.9 (0.6–53.0) 0.9 (0.3–2.3) 2.1 (1.0–4.8)
Fattovich (2001) [60]
Italy
Cohort
29
226
LC
1.1 (0.5–2.3)
Roffi (2001) [13]
Italy
Case-control
62
2245
CH
3.2 (1.8–5.0)
Hayashi (2002) [53]
Japan
Cohort
21
392
CH
1.0 (0.4–2.9)
Ikeda (2002) [61]
Japan
245
348
LC
1.4 (1.0–1.9)
Ohata (2003) [54]
Japan
Retrospective cohort Cohort
ns
Cohort: 161 patients
CH and LC
1.2 (0.4–3.5)
Lee (2006) [1]
Taiwan
Case-control
112
264
Yu (2006) [58]
Taiwan
Retrospective cohort
41
735
CH and LC LC only CH and LC
4.1 (1.8–9.4) 5.7 (2.0–15.8) 2.5 (1.2–5.0)
Bruno (2007) [11]
Italy
Cohort
55
108
LC
3.0 (1.4–6.5)
Kumar (2007) [5]
India
Case-control
21
5
No liver disease
0.7 (0.1–8.2)
(1.2–57.5) (1.1–2.9) (1.0–13.9) (1.2–5.8)
Age, gender, nationality, HBV, etiology, follow-up duration Age, gender, HCV duration, alcohol, IFN, LC at entry, intravenous drugs, ALT, bilirubin Age, gender Age, gender, years since LC diagnosis, department, hospitalization, serum albumin, AFP, AST Age Age, gender Age, gender, date and hospital of admission Age, gender, bilirubin, albumin, PLT, therapy, biochemical remission Multivariate estimate (not otherwise specified) Age, gender, PLT, response to IFN, HCV RNA, ALT, staging, grading Age, gender, albumin, AFP, PLT Age, gender, diabetes, BMI, alcohol, ALT, HCV core protein, IFN, LC, grading, steatosis Age, HCV RNA levels Age, gender, LC, PLT, AST, APRI, AFP, IFN, response to IFN Age, gender, child class, varices, alcohol, diabetes, laboratory values, IFN Age, gender, ethnicity, time period
LC, liver cirrhosis; CH, chronic hepatitis; ALT, alanine transaminase; AFP, a-fetoprotein; AST, asparate aminotransferase; BMI, body mass index; IFN, interferon treatment; PLT, platelet; APRI, AST-to-PLT ratio index. a RR for patients with HCV genotype 1b (or HCV type 1, if information on subtypes was not available) vs. patients with other HCV genotypes. b RR for HCV-1 vs. HCV-2.
Table 2 presents a brief description of the 36 studies (nine cohort and 27 case-control studies) reporting only crude data on HCV genotypes and HCC, with the corresponding RR and 95% CI. These articles were published from 1992 to 2007 and referred to a study period ranging from 1966 to 2003. Twenty studies were conducted in Asia, 11 in Europe, three in Africa, and two in America. Thirty-two studies included patients
with specific CLD (CH or LC) and in 24 of them separate data for LC patients were available. 3.2. Quantitative data synthesis Overall, we found a significant association between HCV genotype 1b and HCC (summary RR 1.78; 95% CI 1.36–2.32) based on the 21 studies providing adjusted
S. Raimondi et al. / Journal of Hepatology 50 (2009) 1142–1154
1145
Table 2 Description of the 36 studies presenting only crude data on HCV genotype 1b and HCC development. Author (year)
Country
Study design
N cases
N controls
Study population
RR (95% CI)a
Okamoto (1992) [76]
Japan
Case-control
16
230
0.32 (0.12–0.90)
Chen (1994) [77] Ichimura (1994) [31]
Taiwan Japan
Case-control Case-control
32 48
25 269
Shrestha (1994) [78] Tanaka (1994) [79] Nousbaum (1995) [41]
Case-control Case-control Case-control
2 94 33
5 187 149
Takano (1995) [4] Yamada (1995) [80] Yotsuyanagi (1995) [15] Kobayashi (1996) [73] Lee (1996) [81]
Nepal Japan France and Italy Japan Japan Japan Japan Korea
HCV liver disease and no liver diseaseb,c LC HCV liver disease and no liver diseaseb,c CH and LCc CH and LCc CH and LCc
Cohort Case-control Case-control Cohort Case-control
9 604 67 31 28
84 3225 129 105 108
Sithebe (1996) [82] Soetjipto (1996) [83] Takada (1996) [14] Benvegnu` (1997) [17] Fernandez (1997) [84] Han (1997) [18]
South Africa Indonesia Japan Italy Spain Korea
Case-control Case-control Case-control Cohort Case-control Case-control
1 21 401 15 29 37
7 58 2194 346 88 128
Haydon (1997) [85] Mangia (1997) [74] Romeo (1997) [86] Shah (1997) [87] Shimizu (1997) [88] Bellentani (1999) [89] Bonis (1999) [68] Giannini (1999) [67] Reid (1999) [70] Matsumura (2000) [69] Tanaka (2000) [90] Widell (2000) [91] Zekri (2000) [92] Franchini (2001) [93] Yu (2001) [94] Liu (2003) [95] Shiratori (2005) [96] Abdel-Hamid (2007) [97]
Scotland Italy Italy Pakistan China Italy USA Italy USA Japan Japan Sweden Egypt Italy Taiwan Taiwan Japan Egypt
Case-control Case-control Cohort Case-control Case-control Cohort Cohort Case-control Case-control Case-control Case-control Case-control Case-control Cohort Case-control Cohort Cohort Case-control
20 76 56 7 10 5 10 39 13 55 7 25 29 1 61 3 119 28
28 117 123 38 22 20 145 223 33 472 48 53 46 81 591 7 226 93
Stankovic-Djordjevic (2007) [98]
Serbia
Case-control
36
35
CH CH and LCc CH and LCc CH HCV liver disease and no liver diseaseb,c HCV liver disease No liver disease CH and LCc CH and LCc CH and LC CH, LC, and no liver diseaseb,c LC CH LC HCV liver diseaseb HCV liver diseaseb,c LC CH and LC CH and LCc LC CH and LCc HCV liver diseaseb,c No liver disease CH and no liver diseaseb CH, LC, and no liver diseaseb,c HCV liver diseaseb LC LC HCV liver disease and no liver disease LC
3.76 (0.86–16.43) 1.61 (0.76–3.38) 0.25 (0.01–8.56) 1.38 (0.69–2.77) 2.13 (0.77–5.88) 1.17 1.46 1.58 6.94 0.63
(0.22–6.06) (1.17–1.83) (0.72–3.49) (1.56–30.81) (0.26–1.53)
45.00 (0.61–3297.13) 4.19 (1.46–12.01) 1.52 (1.18–1.97) 1.31 (0.45–3.75) 11.48 (0.66–198.78) 0.77 (0.37–1.62) 10.00 (2.54–39.29) 1.33 (0.74–2.39) 0.89 (0.47–1.68) 0.51 (0.02–10.54) 0.19 (0.02–2.40) 0.29 (0.04–2.17) 3.40 (0.85–13.69) 1.80 (0.90–3.59) 6.00 (1.47–24.45) 0.92 (0.48–1.75) 4.60 (0.24–86.84) 7.12 (2.33–21.76) 8.45 (0.39–182.64) 5.07 (0.20–128.29) 2.54 (1.40–4.61) 1.50 (0.09–25.39) 0.88 (0.57–1.36) 7.08 (0.62–81.17) 4.16 (1.48–11.64)
RR, Relative Risk; CI, Confidence Interval; LC, liver cirrhosis; CH, chronic hepatitis. a RR for patients with HCV genotype 1b (or HCV type 1, if information on subtypes was not available) vs. patients with other HCV genotypes. b Separate estimate for CLD patients was available. c Separate estimate for LC patients was available.
RR (Table 3 and Fig. 1). The summary RR slightly increased when we excluded papers presenting estimates for HCV type 1, without further subtype differentiation (RR 2.01; 95% CI 1.53–2.66) and when we compared HCV genotype 1b with HCV type 2 (RR 2.17; 95% CI 1.50–3.15). Pooled RR was significantly lower for studies with no subtype specification than for studies which specifically investigated HCV genotype 1b (meta-regression p-value: 0.04). We observed similar results when the
analysis was restricted to the papers including only CLD patients (Table 3). Among the eight studies including only LC patients, the RR (95% CI) for the association between HCV genotype 1b and HCC was still significant: 1.60 (1.07–2.39) (Table 3 and Fig. 2). After exclusion of a single study providing risk estimate for HCV type 1, the association remained of borderline statistical significance but was no more significant when we compared HCV genotype 1b with HCV type 2, even if it
1146
S. Raimondi et al. / Journal of Hepatology 50 (2009) 1142–1154
Table 3 Summary Relative Risk (RR) and 95% Confidence Interval (CI) for the association between HCV genotypes and HCC development, and heterogeneity estimates in the 21 studies presenting adjusted RR. Study population
HCV genotype comparison a
N studies
N cases/N controls
RR (95% CI)
Q test p-value (I2)
Any HCV positive
1b/1 vs. non 1b/1 1b vs. non 1b 1b vs. 2
21 17 10
1408/6677 1128/5773 797/1821
1.78 (1.36–2.32) 2.01 (1.53–2.66) 2.17 (1.50–3.15)
0.10 (47%) 0.06 (38%) 0.11 (37%)
Chronic Liver Diseaseb
1b/1a vs. non 1b/1 1b vs. non 1b 1b vs. 2
16 14 7
1016/5096 802/4747 440/796
1.80 (1.35–2.40) 1.92 (1.37–2.67) 1.94 (1.13–3.32)
0.02 (48%) 0.02 (48%) 0.03 (56%)
Liver Cirrhosis
1b/1a vs. non 1b/1 1b vs. non 1b 1b vs. 2
8 7 4
596/935 463/851 222/413
1.60 (1.07–2.39) 1.66 (0.98–2.82) 1.80 (0.58–5.62)
0.04 (52%) 0.03 (57%) 0.01 (72%)
a b
HCV type 1 was considered ‘‘at risk” genotype when information on HCV genotype 1b was not available. Includes chronic hepatitis and liver cirrhosis.
was higher (RR 1.80; 95% CI 0.58–5.62). On the contrary, when we considered only the seven studies excluding patients with LC, the RR (95% CI) increased to 2.46 (1.69–3.59). The results from the whole set of studies providing only crude data were shown in Fig. 3 and were similar to those reported for the 21 studies providing adjusted risk estimates: the RRs (95% CI) were 1.63 (1.30–2.06) and 1.51 (1.23–1.87) for any HCV positive subjects and CLD patients, respectively. When only LC patients were included, the association lost statistical significance (RR 1.13; 95% CI 0.89–1.44). However, pooling together the 32 studies on LC patients (eight providing adjusted estimate and 24 only crude data), we found a still significant association between HCV genotype 1b and HCC (RR 1.28; 95% CI 1.03–1.58). 3.3. Heterogeneity and publication bias Pooled RR of studies including only LC patients was significantly lower than that of studies excluding LC patients (meta-regression p-value: 0.04). We observed a significant heterogeneity among the 21 studies providing adjusted risk estimate and in stratifications by study population or HCV genotype comparison group, as shown in Table 3. Study design seemed to explain the largest part of heterogeneity both when all studies where combined and when articles with CLD patients were included. Cohort studies presented lower RR in comparison to case-control studies: for the whole set of 21 studies RRs (95% CI) were 1.27 (0.87–1.86) and 2.41 (1.82– 3.18) for cohort and case-control studies, respectively, with a significant meta-regression p-value of 0.01 (Table 4). Among the 12 case-control studies, the hypothesis of homogeneity could be accepted, with Q test pvalue = 0.34 and I2 = 10%. Among the nine cohort studies, heterogeneity was still observed (Q test pvalue = 0.06, I2 = 50%) but it could be easily explained by the fact that earlier studies presented significantly
lower estimates than more recent ones, as clearly shown in Fig. 1 (meta-regression p-value for publication year: