Fibrosis-stage Specific Incidence of Hepatocellular Cancer after Hepatitis C Cure with Direct-Acting Antivirals: A Systematic Review & Meta-analysis

Abstract

Background and Aims:

Hepatitis C virus (HCV) eradication with direct-acting antivirals (DAA) reduces hepatocellular carcinoma (HCC) risk. Pooled HCC incidence rates by cirrhosis status and fibrosis stage have not been estimated using meta-analysis.

Methods:

We searched PubMed, Web of Science, Embase, and Cochrane Library from 1/1/2014–12/31/2020 to identify studies assessing HCC incidence or outcomes by cirrhosis status, in adults with HCV who achieved sustained virologic response (SVR) after DAAs. Pooled estimates were obtained using random-effects modeling. Subgroup, sensitivity, and meta-regression analyses were performed to evaluate heterogeneity.

Results:

We included 31 studies involving 27,711 patients with cirrhosis (mean follow-up 2.1 years) and 11 studies involving 32,123 patients without cirrhosis (mean follow-up 2.6 years). HCC incidence was 2.99/100 person-years (95% CI: 2.52–3.54; I²=75%) in patients with cirrhosis, 0.47/100 person-years (95% CI: 0.32–0.70, I²=71%) in patients without cirrhosis, and 0.63/100 person-years (95% CI: 0.34–1.20, I²=0%) in F3 fibrosis. Among cirrhosis patients, HCC incidence was highest in studies with <1 year of follow-up (6.17/100 person-years [95% CI: 3.73–10.19]) and progressively lower in studies with longer follow-up (1–2 years: 2.75/100 person-years [95% CI: 2.48–3.06], 2–3 years: 2.90/100 person-years [95% CI: 1.90–4.44], ≥3 years: 1.83/100 person-years [95% CI: 0.88–3.80]).

Conclusion:

Pooled HCC incidence after SVR in patients with cirrhosis was very high (2.99/100 person-years), but may be declining as longer time accrues after SVR. In patients without cirrhosis, including F3 fibrosis, HCC incidence was lower than thresholds associated with cost-effective HCC screening. In patients with F3 fibrosis, the lack of between-study heterogeneity provides strong evidence that HCC screening may not be warranted.

Graphical Abstract

INTRODUCTION

Chronic hepatitis C virus (HCV) infection affects over 71 million people worldwide.¹ It is a major cause of liver cirrhosis and hepatocellular carcinoma (HCC), accounting for 290,000 deaths in 2019.² Interferon-based HCV treatments used to be the standard of care, but were limited by significant side effects, long treatment duration, and low effectiveness. Since 2014, highly effective, safe and well-tolerated oral direct-acting antiviral (DAA) therapies have transformed HCV therapy. Approximately 9.4 million people are believed to have received DAA-based HCV treatment worldwide between 2015–2019.²

DAA-induced HCV cure results in significant reduction in HCC risk in patients with and without cirrhosis.^3–6 However, the actual incidence of HCC after DAA-induced HCV eradication (known as sustained virologic response or SVR) in patients with and without cirrhosis, and according to fibrosis stage, is unclear. Professional liver societies also differ in their HCC screening recommendations after HCV cure. The American Association for the Study of Liver Diseases (AASLD) recommends HCC screening only in patients with pre-treatment cirrhosis.⁷ The European Association for the Study of the Liver (EASL) recommends HCC screening in patients with pre-treatment F3 fibrosis and cirrhosis.⁸ The Asian Pacific Association for the Study of the Liver (APASL) recommends HCC screening in all patients after SVR, irrespective of fibrosis stage and presence of cirrhosis.⁹ Obtaining accurate estimates of HCC incidence after DAA-induced eradication is critical because HCC incidence is positively correlated with the potential benefits of screening, and is the main determinant of the cost-effectiveness of HCC screening after HCV eradication.¹⁰ We therefore performed a systematic review and meta-analysis on available literature to summarize HCC incidence and outcomes in patients with and without cirrhosis, who achieved SVR after DAA therapy.

METHODS

SEARCH STRATEGY AND SELECTION CRITERIA

We performed a literature search on PubMed, Web of Science, Embase, and Cochrane Library databases for English-language and human participant studies that were published from 1/1/2014 to 12/31/2020. We used the following search terms: (hepatitis C OR HCV) AND (cirrhosis) AND (direct acting antiviral* OR DAA OR sofosbuvir OR interferon-free) AND (hepatocellular carcinoma OR HCC OR liver cancer) AND (incidence OR occurrence OR stage* OR treatment* OR mortality OR survival OR outcome*). Articles were included if they assessed HCC incidence or outcomes by cirrhosis status in adult patients with a history of HCV and receipt of DAAs. Studies prior to 2014 were excluded as sofosbuvir was approved by the U.S. Food and Drug Administration and European Medicines Agency for use in late 2013. We also excluded: (1) studies on pediatric populations (age <18 years); (2) review articles, editorials, conference abstracts, and non-English language articles (as the methods could not be assessed in detail); (3) studies only assessing HCC recurrence; (4) studies of patients who received only interferon-based therapy; and (5) studies of patients who received DAA therapy that did not provide data by SVR and cirrhosis status.

Two authors (NJK, PV) independently reviewed the title and abstract of all articles to select relevant studies for initial inclusion and full-text review. Data extraction was performed using a pre-determined standardized form by two independent reviewers (NJK, PV, or AC) and in the case of discrepancies, a third reviewer (EC or NJK) provided arbitration. We extracted the following data from studies that met full-text inclusion criteria: (1) study characteristics (first author’s last name, year, country, study design, HCC screening program before and after treatment); (2) patient characteristics (number of patients, age, race/ethnicity, DAA regimen, prior interferon-based treatment, HCV genotype, presence of cirrhosis, fibrosis stage, Child-Pugh score, duration of follow-up); (3) HCC characteristics (incidence, number of lesions, Barcelona Clinic Liver Cancer [BCLC] stage, Milan criteria, or radiologic staging); (4) HCC treatment modality (surgical resection, ablation [microwave ablation, radiofrequency ablation, cryoablation, percutaneous ethanol injection], transarterial therapies [transarterial embolization, transarterial chemoembolization, selective internal radiation therapy including yittrium-90], systemic therapy [chemotherapy, immunotherapy], liver transplantation, and/or supportive/palliative care); and (5) other outcomes of interest (all-cause and liver-related mortality). We also reviewed study population descriptions when possible, to exclude potentially overlapping cohorts by country and/or time. Article screening and data-extraction were performed using Covidence software (Veritas Health Innovation, Melbourne, Australia; www.covidence.org). Data collection was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.¹¹ This study protocol was registered in the PROSPERO database (CRD42021235023).

STUDY DEFINITIONS

We defined cirrhosis using standard histological, radiological, and/or clinical criteria. SVR was defined as documentation of a negative HCV viral load test ≥12 weeks after completion of DAA therapy. The presence of an adequate HCC screening program after treatment initiation was defined as a study description of imaging (ultrasound, dynamic contrast-enhanced computed tomography [CT] or magnetic resonance imaging [MRI]) with or without serum alpha-fetoprotein, at least every 6 months per AASLD guidelines¹². HCC incidence was defined as the new diagnosis of primary liver cancer in patients without a prior diagnosis of HCC. HCC staging was determined using the BCLC staging system in accordance with existing practice guidelines.^{7, 8, 13} HCC treatment was categorized as surgical resection, ablation, transarterial therapies, systemic therapy, liver transplantation, and/or supportive/palliative care. Potentially curative therapies included resection, ablation, and liver transplantation. The presence of hepatitis B virus (HBV) or human immunodeficiency virus (HIV) co-infection was defined by study inclusion and exclusion criteria. Mortality was categorized as all-cause or liver-specific mortality.

STUDY OUTCOMES

The primary outcome was HCC incidence following SVR after DAA therapy. Secondary outcomes were all-cause and liver-related mortality, HCC staging at diagnosis, and HCC treatment modality. We analyzed these outcomes according to cirrhosis status (cirrhosis or no cirrhosis) and where available, fibrosis stage (F0, F1, F2, F3).

STUDY QUALITY ASSESSMENT

We used the Newcastle-Ottawa Quality Assessment scale (NOS)¹⁴ to assess the study quality of articles. Two reviewers (NJK, PV, or AC) independently scored each study based on selection, comparability, outcome, and exposure, with a maximum score of 9. In the case of discrepancies, a third reviewer (EC or NJK) provided arbitration. Studies were categorized as good (score ≥7), fair (score 5–6), or poor (score <5).¹⁵ We also assessed for differences in study definitions for cirrhosis, F3 fibrosis, and pre-/post-treatment HCC screening.

DATA ANALYSIS

We used random-effects models with a 95% confidence interval (CI) to obtain pooled estimates of HCC incidence, and all-cause and liver-related mortality rates calculated per 100 person-years by cirrhosis status. To measure between-study heterogeneity, we used the I² statistic with I²≥50% suggesting substantial heterogeneity. We further investigated heterogeneity using several methods. We performed subgroup analyses based on geographic region, study design (prospective vs. retrospective), duration of study follow-up, study quality, presence of an HCC screening program before and after treatment, and HBV or HIV co-infection. We then identified studies with CIs that did not overlap with the pooled effect to perform an outlier analysis. Lastly, we used meta-regression accounting for geographic region, study follow-up time, publication year, HCC screening program before and after treatment, study design and quality, and viral co-infection, as applicable. Descriptive analyses were performed to summarize HCC staging and characteristics at diagnosis for patients with cirrhosis. Publication bias was assessed using a funnel plot and Egger’s regression test. R version 4.1.0 (Vienna, Austria) was used for analysis.

RESULTS

LITERATURE SEARCH

After removal of duplicates, the literature search yielded 2,461 articles. We reviewed 152 articles in full-text and performed data extraction for 31 articles^{5, 16–45} that included patients with cirrhosis and 11 articles^{19, 21, 22, 28–30, 34, 35, 43–45} that included patients without cirrhosis. The screening and article inclusion/exclusion process are shown in Figure 1.

Figure 1. PRISMA flow diagram of study selection.

HCC: hepatocellular carcinoma; SVR: sustained virologic response.

STUDY AND PATIENT CHARACTERISTICS

The 31 studies extracted for patients with cirrhosis included 27,711 patients (Table 1). Among the 15 studies where information on sex was provided by cirrhosis status, 97% of patients (n=10,855/27,711) were male. Sixteen articles originated from Europe^{5, 16, 17, 19, 20, 22, 24–26, 28, 31, 32, 34, 36, 41, 42,} articles from Asia^{18, 29, 35, 40, 44}, 3 articles from North America^{21, 23, 30,} 5 articles from the Middle East^{27, 37, 38, 43, 45,} 1 article from South America³³, and 1 article from a mixed European/North American cohort³⁹. The definition of cirrhosis varied across studies (Table 2). Cirrhosis was defined by FIB-4 score ≥3.25 in 1 study⁴⁰, liver stiffness measurement (LSM) ≥10 kPa in 1 study⁴¹, LSM ≥14 kPa in 1 study²⁴, ICD 9 and 10 codes in 1 study³⁰, histology in 1 study²⁶, clinical criteria in 2 studies^{16, 18}, and a mixture of clinical, histological, LSM, or biochemical criteria in 21 studies5, ^{17, 19, 21–23, 25, 28, 29, 31–37, 39, 42–45}. The definition of cirrhosis was not clearly defined in 3 of the studies^{20, 27, 38}.

Table 1.

Characteristics of the 31 studies included in the meta-analysis of studies reporting hepatocellular carcinoma incidence following DAA-induced SVR

	Overall study			Patients with cirrhosis			Patients without cirrhosis
First author (Year/Country)	Study design	Period of inclusiona	Total N	N	Age, years	Sex (% male)	N (F3)	Age, years	Sex (% male)
Cheung (2016/United Kingdom)	P	4/1/2014–11/11/2014	406	288	NS	NS	-	-	-
Conti (2016/Italy)	R	3/2015–11/2015	344	261	NS	NS	-	-	-
Ji (2017/China)	P	4/2015–12/2015	30	27	NS	NS	-	-	-
Calvaruso (2018/Italy)	P	3/1/2015–11/30/2016	2249	2140	NS	NS	-	-	-
Finkelmeier (2018/Germany)	P	1/2014–8/2016	819	242	NS	NS	522	NS	NS
Innes (2018/Scotland)	R	1/1/1997–4/1/2016	857	272	52.1 (mean)	194 (71.3)	-	-	-
Kim (2018/USA)	R	1/2014–1/2016	192	55	NS	NS	128 (15)	57 (median)	64 (56.6)
Kozbial (2018/Austria)	R	10/20134/2016	551	393	58.2 (mean)	245 (62.3)	158 (158)	57.3 (mean)	92 (58.2)
Kuftinec (2018/USA)	R	1/1/2014–4/1/2017	150	140	NS	NS	-	-	-
Merchante (2018/Spain)	P	2/2006–3/31/2017	495	224	NS	NS	-	-	-
Mettke (2018/Germany)	P	1/20/2014–2/23/2016	158	158	59 (median)	87 (55.1)	-	-	-
Nahon (2018/France)	P	2006–2012	1270	274	59.2 (median)	173 (63.1)	-	-	-
Aziz (2019/Pakistan)	R	6/2016–1/2018	300	276	NS	163 (59.1)	-	-	-
Carrat (2019/France)	P	8/6/2012–12/31/2015	9895	2329	NS	NS	3286	NS	NS
Ide (2019/Japan)	P	1/1/2015–1/31/2017	2552	648	NS	NS	1904	NS	NS
Ioannou (2019/USA)	R	1/1/2000–12/31/2015	4813 5	7533	61.9 (mean)	7307 (97.0)	21500	60.8 (mean)	20640 (96.0)
Lleo (2019/Italy)	P	1/2015–12/2015	1927	1679	NS	NS	-	-	-
Marino (2019/Spain)	R	Before 10/30/2015	1119	1070	NS	NS	-	-	-
Pinero (2019/Argentina/Uruguay/Chile/Columbia/Brazil)	P	5/1/2016–6/1/2018	1400	653	NS	NS	-	-	-
Rinaldi (2019/Italy)	P	2/2015–2/2017	985	714	NS	NS	252 (252)	NS	NS
Abe (2020/Japan)	R	12/2014–4/2019	1068	188	70 (median)	90 (47.9)	880	66 (median)	421 (47.8)
Campello (2020/Italy)	P	9/2015–3/2016	58	58	59 (median)	38 (65.5)	-	-	-
Gogichaishvili (2020/Georgia)	R	4/2015–3/2016	408	401	55 (median)	312 (77.8)	-	-	-
Hassany (2020/Egypt)	P	3/2016–3/2019	350	329	58.2 (mean)	178 (54.1)	-	-	-
Krassenburg (2020/Netherlands/Germany/Canada)	R	1/1/2015–7/1/2015	868	766	NS	479 (62.5)	-	-	-
Nagaoki (2020/Japan)	R	1/1998–1/2019	383	173	73 (median)	70 (40.5)	-	-	-
Pons (2020/Spain)	P	1/1/2015–3/31/2016	572	572	63.7 (mean)	282 (49.3)	-	-	-
Sangiovanni (2020/Italy)	P	6/2015–6/2017	1285	1119	NS	NS	-	-	-
Shiha (2020/Egypt)	P	1/2015–8/2017	2372	1734	NS	NS	638 (638)	NS	NS
Tanaka (2020/Hong Kong/Japan/South Korea/Taiwan)	R	2014–2018	5814	2911	NS	1194 (41.0)	2735	NS	1210 (44.2)
Yousif (2020/Egypt)	P	1/2017–6/2018	208	84	56 (mean)	43 (51.2)	120	NS	NS

^aRefers to time of treatment or study recruitment (in Nahon 2018, patients were followed until 12/2016).

For studies pre-dating 2014, only data pertinent to DAA therapy (after 2014) was included in the meta-analysis. DAA: direct-acting antivirals; NS: Not specified by cirrhosis status; P: prospective; R: retrospective; SVR: sustained virologic response.

Table 2.

Quality assessment of the 31 studies included in the meta-analysis of studies reporting hepatocellular carcinoma incidence following DAA-induced SVR

First author (Year)	NOSa (max score 9, category)	Definition of cirrhosis	Definition of F3 fibrosis	Pre-treatment HCC evaluation	HCC screening program
Cheung (2016)	6 (fair)	Clinical	-	Not specified	None
Conti (2016)	8 (good)	Metavir F4 or LSM ≥12 kPa	-	US before treatment	US every 3 months for 6 months after treatment
Ji (2017)	6 (fair)	Clinical	-	US/CT + AFP before treatment	US every 3 months
Calvaruso (2018)	8 (good)	Metavir F4, LSM >12 kPa, platelet <100k, or varices	-	US before enrollment (range 1–36 months before)	US every 6–12 months during/after treatment
Finkelmeier (2018)	8 (good)	LSM ≥12.5 kPa, histology, clinical, biochemical, imaging, or endoscopy	LSM >9.5 to <12.5 kPa	US/CT/MRI within 4 weeks before treatment	None
Innes (2018)	8 (good)	Not specified	-	Not specified	US every 6 months
Kim (2018)	7 (good)	Imaging, Metavir F4 or FibroTest F4	Metavir F3 or FibroTest F3	Not specified	None
Kozbial (2018)	8 (good)	LSM ≥12.5 kPa or clinical	LSM 9.3 to <12.5 kPa	US/CT/MRI up to 6 months before treatment	US every 6 months
Kuftinec (2018)	8 (good)	Imaging, clinical, biochemical, or FIB-4	-	US/CT/MRI up to 12 months before treatment	US, CT, or MRI within 12 months
Merchante (2018)	7 (good)	LSM ≥14 kPa	-	Not specified	US every 6 months
Mettke (2018)	7 (good)	Ishak F5–6, LSM ≥14.5 kPa, imaging, or clinical	-	US/CT/MRI at study enrollment	US every 6 months
Nahon (2018)	8 (good)	Histology	-	US prior to study enrollment	US every 6 months
Aziz (2019)	7 (good)	Not specified	-	US at study enrollment	US every month during treatment, US every 3 months for 6 months after treatment
Carrat (2019)	8 (good)	LSM >12.5 kPa, Metavir F4, FibroTest >0.75, Fibrometer >0.98, Hepascore >0.84, platelet <150k, or PT <70%	Metavir F3, LSM 9.5–12.5 kPa, FibroTest 0.59–0.75, Fibrometer 0.62- 0.98, or Hepascore 0.610.84	Not specified	None
Ide (2019)	8 (good)	High FIB-4, biochemical, imaging, or clinical	Not specified	US/CT/MRI at study enrollment	None
Ioannou (2019)	8 (good)	ICD 9/10 codes at least twice	Not specified	Excluded HCC cases diagnosed <180 days of treatment initiation	None
Lleo (2019)	8 (good)	Metavir F4, varices, or LSM >12.5 kPa	-	US + AFP within 3 months of treatment initiation	US + AFP every 6 months
Marino (2019)	8 (good)	LSM ≥14 kPa, histology, clinical, imaging	-	Abdominal imaging before study enrollment	US (frequency unknown)
Pinero (2019)	8 (good)	LSM ≥12.5 kPa, clinical, histology, or FibroTest	-	HCC surveillance prior to treatment initiation	US ± AFP every 6 months
Rinaldi (2019)	7 (good)	Clinical, biochemical, imaging, histology, or elastography	Metavir or LSM F3	HCC screening before at study enrollment	US every 6 months
Abe (2020)	8 (good)	Imaging or histology	Not specified	Excluded HCC cases diagnosed <180 days of treatment initiation	US + AFP every 6–12 months during and after treatment
Campello (2020)	8 (good)	Clinical, biochemical, or imaging	-	Not specified	None
Gogichaishvili (2020)	7 (good)	Metavir F4 or clinical	-	Not specified	US every 6 months
Hassany (2020)	8 (good)	Not specified	-	US before treatment	AFP at 3 months, US every 4 months
Krassenburg (2020)	7 (good)	LSM ≥13 kPa, Metavir F4, or 2+ of: clinical, biochemical, varices, imaging, serum panel F4	-	Medical chart review before treatment	None
Nagaoki (2020)	7 (good)	FIB-4 ≥3.25	-	Not specified	AFP, US, and/or CT every 6 months
Pons (2020)	7 (good)	LSM ≥10 kPa	-	US within 6 months of treatment initiation	US every 6 months
Sangiovanni (2020)	8 (good)	Clinical, LSM >12 kPa, Metavir F4, or Ishak F5–6	-	Date of last US before study documented	US every 6 months
Shiha (2020)	8 (good)	Clinical, imaging, or LSM >16.3 kPa	LSM 10.2–16.3 kPa	US before treatment initiation	US every 6 months
Tanaka (2020)	8 (good)	Histology, LSM >12.5 kPa, FIB-4 >3.25, platelet <120k, clinical, or endoscopy	Not specified	Excluded HCC cases diagnosed <180 days of treatment initiation	US + AFP every 6 months
Yousif (2020)	7 (good)	Clinical, LSM >12 kPa, FIB-4 >3.25 or APRI ≥2, and imaging	FIB-4, APRI, LSM, and US	US done before study enrollment	None

AFP: alpha-fetoprotein; APRI: AST to platelet ratio index; CT: computed tomography; DAA: direct-acting antivirals; FIB-4: Fibrosis-4 score; LSM: liver stiffness measurement; MRI: magnetic resonance imaging; NOS: Newcastle-Ottawa Quality Assessment scale. PT: Prothrombin time; SVR: sustained virologic response; US: ultrasound.

The 11 studies extracted for patients without cirrhosis included 32,123 patients (1,063 with F3 fibrosis, 31,060 with unspecified fibrosis stage) (Table 1). Among the 5 studies where information on sex was provided by cirrhosis status, 88% of patients (n=22,427/25,386) were male. Four articles originated from Europe^{19, 22, 28, 34}, 2 articles from North America^{21, 30}, 3 articles from Asia^{29, 35, 44}, and 2 articles from the Middle East^{43, 45}. The definition of F3 fibrosis also varied across studies (Table 2). Three studies defined F3 fibrosis by LSM values alone^{19, 22, 43}, with values ranging from 9.3–16.3 kPa. Four studies used a combination of LSM, FIB-4, APRI, imaging, histology, and/or other laboratory-based staging^{21, 28, 34, 45}. The definition of F3 fibrosis was not clearly defined in 4 of the studies^{29, 30, 35, 44}.

Twenty-three studies included a description of pre-treatment HCC evaluation. Twenty-two studies described an HCC screening program after treatment initiation. Additional study and patient characteristics are summarized in Tables 1–2.

QUALITY ASSESSMENT

Quality assessment of articles indicated that 29 studies^{5, 17, 19–45} were of good quality with low risk of bias and 2 studies^{16, 18} were of fair quality with moderate risk of bias. Study definitions used to define cirrhosis, F3 fibrosis, and HCC screening before and after treatment, varied between studies (Table 2)

HCC INCIDENCE IN PATIENTS WITH CIRRHOSIS

Among the 27,711 patients with cirrhosis contributing 58,297 person-years, the pooled HCC incidence was 2.99 per 100 person-years (95% CI: 2.52–3.54; I²=75%, 95% CI: 65–83%) (Figure 2A). HCC incidence was similar across geographic regions, ranging from 2.78 to 3.18 per 100 person-years, P=.10) (Table 3). HCC incidence progressively declined with longer study follow-up time (P=.01); it was highest at 6.17 per 100 person-years (95% CI: 3.73–10.19, I²=73%, 95% CI: 22–90%) in studies with <1 year of follow-up, 2.75 per 100 person-years (95% CI: 2.48–3.06, I²=23%, 95% CI: 0–56%) in studies with 1–2 years of follow-up, 2.90 per 100 person-years (95% CI: 1.90–4.44, I²=86%, 95% CI: 65–94%) in studies with 2–3 years of follow-up, and lowest at 1.83 per 100 person-years (95% CI: 0.80–3.80, I²=83%, 95% CI: 61–92%) in studies with ≥3 years of follow-up. Longer study follow-up time was also significantly associated with lower HCC incidence rates in multivariate meta-regression analysis (all P≤.006, Supplemental Table 1).

Figure 2. Forest plots of hepatocellular carcinoma incidence rates among patients with HCV following DAA-induced SVR by cirrhosis and fibrosis status.

(A) Cirrhosis (B) Without cirrhosis (C) F3 fibrosis. Heterogeneity considered substantial if I² ≥50%. DAA: direct-acting antivirals; HCC: hepatocellular carcinoma; HCV: hepatitis C virus; SVR: sustained virologic response.

Table 3.

Estimated pooled hepatocellular carcinoma incidence rates following DAA-induced SVR among patients with HCV cirrhosis overall and by subgroup analyses

	No. of studies	HCC incidence rate per 100 PYs (95% CI)	τ 2	I2 (95% CI)	Psubgroup
Overall	31	2.99 (2.52–3.54)	0.16	75% (65–83)	-
Geographic region a					.10
North America	3	3.10 (3.19–4.39)	0.04	16% (0–91)
Europe	16	2.78 (1.99–3.87)	0.38	98% (76–90)
Asia	5	2.92 (2.54–3.36)	0.003	5% (0–80)
Middle East	5	3.18 (2.75–3.68)	0.0	0% (0–79)
Study follow-up time b					.01
<1 year	4	6.17 (3.73–10.19)	0.17	73% (22–90)
1–2 years	18	2.75 (2.48–3.06)	0.01	23% (0–56)
2–3 years	4	2.90 (1.90–4.44)	0.16	86% (65–94)
3+ years	5	1.83 (0.88–3.80)	0.62	83% (61–92)
Study quality					.24
Good	29	2.96 (2.48–3.52)	0.17	77% (67–84)
Fair	2	4.06 (2.47	0.0	0% (-)
Study design					.50
Prospective	18	2.81 (2.08–3.80)	0.33	83% (74–89)
Retrospective	13	3.15 (2.79–3.55)	0.02	48% (1–73)
Pre-treatment HCC evaluation c					.30
Yes	21	3.17 (2.60–3.85)	0.15	75% (61–83)
Not specified	10	2.53 (1.73–3.70)	0.28	80% (63–89)
HCC screening program after treatment initiation d					.18
Adequate	18	2.62 (2.11–3.24)	0.14	70% (52–82)
Inadequate	4	2.84 (2.42–3.34)	<0.01	10% (0–86)
No	9	3.90 (2.71–5.61)	0.23	87% (77–93)
HBV/HIV co-infection					<.001
None	10	2.78 (2.20–3.51)	0.10	74% (51–86)
No HIV, possible HBV	3	6.15 (2.55–14.8)	0.41	87% (65–96)
No HBV, possible HIV	2	3.04 (2.35–3.94)	0.0	0% (-)
HIV, possible HBV	1	0.32 (0.12–0.85)	-	-
Unknown	15	2.85 (2.67–3.04)	<0.01	22% (0–57)

τ²: between-study variance. I2: between-study heterogeneity (substantial if I²≥50%). P_subgroup: significant if P<.05.

^a2 studies were excluded from geographic region analysis (1 included data from Europe and North America, 1 was the only study from South America).

^bMean or median follow-up time.

^cReceipt of abdominal imaging (ultrasound, CT, or MRI) before treatment or exclusion of HCC diagnosis <180 days of treatment initiation.

^dAdequate screening: abdominal imaging (ultrasound, CT, or MRI) ± serum alpha-fetoprotein at least every 6 months after treatment initiation. Inadequate screening: imaging occurs less than every 6 months or frequency unknown.

CI: Confidence interval; CT: computed tomography; DAA: direct-acting antivirals; HBV: hepatitis B virus, HCC: hepatocellular carcinoma; HCV: hepatitis C virus; HIV: human immunodeficiency virus; MRI: magnetic resonance imaging; PY: Person-years; SVR: sustained virologic response.

HCC INCIDENCE IN PATIENTS WITHOUT CIRRHOSIS

Among the 32,123 patients without cirrhosis contributing 84,474 person-years, the pooled HCC incidence was 0.47 per 100 person-years (95% CI: 0.32–0.70; I²=71%, 95% CI: 46–84%) (Figure 2B). HCC incidence varied by geographic region, ranging from 0.20 per 100 person-years (95% CI: 0.11–0.37, I²=0, 95% CI: 0–85%) in Europe to 0.70 per 100 person-years (95% CI: 0.36–1.34, I²=0%, 95% CI not measurable due to only 2 studies) in the Middle East (P=.03, Table 4). HCC incidence did not differ significantly by study follow-up time (P=.25). Patients with F3 fibrosis (4 studies) had a higher pooled HCC incidence at 0.63 per 100 person-years (95% CI: 0.34–1.20; I²=0%, 95% CI: 0–85%) (Figure 2C) than patients with unspecified fibrosis stage (8 studies) at 0.46 per 100 person-years (95% CI: 0.29–0.73; I²=79%, 95% CI: 58–89%).

Table 4.

Estimated pooled hepatocellular carcinoma incidence rates following DAA-induced SVR among patients without HCV cirrhosis overall and by subgroup analyses

	No. of studies	HCC incidence rate per 100 PYs (95% CI)	τ 2	I2 (95% CI)	P subgroup
Overall	11	0.47 (0.32–0.70)	0.22	71% (46–84)	-
Geographic region					.03
North America	2	0.45 (0.40–0.51)	0.0	0% (-)
Europe	4	0.20 (0.11–0.37)	0.0	0% (0–85)
Asia	3	0.59 (0.30–1.14)	0.30	89% (70–96)
Middle East	2	0.70 (0.36–1.34)	0.0	0% (-)
Study follow-up time a					.25
<1 year	2	0.17 (0.02–1.20)	0.0	0% (-)
1–2 years	6	0.52 (0.26–1.06)	0.41	73% (39–88)
2–3 years	2	0.39 (0.28–0.56)	0.05	67% (0–93)
Study design					.78
Prospective	6	0.49 (0.24–1.03)	0.44	75% (42–89)
Retrospective	5	0.44 (0.31–0.62)	0.06	36% (0–76)
Fibrosis stage b					.42
F3	4	0.63 (0.34–1.20)	0.0	0% (0–85)
Unspecified	8	0.46 (0.29–0.73)	0.26	79% (58–89)
Pre-treatment HCC evaluation c					.01
Yes	9	0.54 (0.38–0.78)	0.15	71% (42–85)
Not specified	2	0.21 (0.11–0.39)	0.0	0% (-)
HCC screening program after treatment initiation d					.41
Adequate	4	0.40 (0.22–0.74)	0.14	17% (0–87)
Inadequate	1	0.65 (0.42–1.01)	-	-
No	6	0.45 (0.23–0.89)	0.40	81% (60–91)
HBV/HIV co-infection					.003
None	6	0.36 (0.21–0.61)	0.16	33% (0–73)
No HIV, possible HBV	1	0.13 (0.01–2.13)	-	-
No HBV, possible HIV	1	0.98 (0.70–1.36)	-	-
Unknown	3	0.50 (0.37–0.68)	0.03	27% (0–92)

τ2: between-study variance. I2: between-study heterogeneity (substantial if I²≥50%). P_subgroup: significant if P<.05.

^aMean or median follow-up time. 1 study was excluded from analysis, as it was the only study with ≥3 years of follow-up.

^b12 studies were included in fibrosis stage analysis as 1 study included both F3 and non-F3 patients. Separate subgroup analysis for F0-F2 was not performed as only 1 study specified F0-F2 stage.

^cReceipt of abdominal imaging (ultrasound, CT, or MRI) before treatment or exclusion of HCC diagnosis <180 days of treatment initiation.

^dAdequate screening: abdominal imaging (ultrasound, CT, or MRI) ± serum alpha-fetoprotein at least every 6 months after treatment initiation. Inadequate screening: imaging occurs less than every 6 months or frequency unknown.

CI: Confidence interval; CT: computed tomography; DAA: direct-acting antivirals; HBV: hepatitis B virus; HCC: hepatocellular carcinoma; HCV: hepatitis C virus; MRI: magnetic resonance imaging; PY: HIV: human immunodeficiency virus; Person-years; SVR: sustained virologic response

HCC STAGING AND TREATMENT

Of the 109 HCC cases with data on tumor burden (5 studies; 4 prospective, 4 with post-treatment HCC screening), 70% (n=76) presented as a single lesion, 29% (n=32) had ≥2 lesions, and 1% (n=1) did not have data available. Of 106 HCC cases with data on BCLC staging (4 studies), 0% (n=0) presented at stage 0, 73% (n=77) at stage A, 8.5% (n=9) at stage B, 3.8% (n=4) at stage C, 1.9% (n=2) at stage D, and 13.2% (n=14) with unspecified staging. Of 33 HCC cases with data on treatment (3 studies), modalities included 12% (n=4) surgical resection, 36% (n=12) ablation, 30% (n=10) transarterial therapies, 6% (n=2) systemic therapy, 9% (n=3) liver transplantation, 12% (n=4) supportive/palliative care, and 3% (n=1) was lost to follow-up. Two patients received transarterial chemoembolization with liver transplantation and one patient received transarterial chemoembolization with radiofrequency ablation. Overall, 60.6% (n=20) received potentially curative therapies.

MORTALITY BY CIRRHOSIS STATUS

Among patients with cirrhosis, the pooled all-cause mortality (9 studies) was 1.18 per 100 person-years (95% CI: 0.65–2.15; I²=80%, 95% CI: 64–89%) and the liver-related mortality (4 studies) was 0.46 per 100 person-years (95% CI: 0.24–0.89; I²=50%, 95% CI: 0–84%) (Supplemental Figure 1A/B). Among patients without cirrhosis, the pooled all-cause mortality (4 studies) was 0.46 per 100 person-years (95% CI: 0.33–0.64; I²=0%, 95% CI: 0–85%) and the liver-related mortality (2 studies) was 0.10 per 100 person-years (95% CI: 0.05–0.20; I²=0%, 95% CI not measurable due to only 2 studies) (Supplemental Figure 1C/D).

HETEROGENEITY INVESTIGATION

We found substantial between-study heterogeneity (defined as I²≥50%) by cirrhosis status and subgroup analyses (Table 3, 4). To further investigate this heterogeneity, we first performed an outlier analysis. In studies on patients with cirrhosis, the removal of 3 outlier studies^{19, 24, 41} resulted in an HCC incidence of 3.01 per 100 person-years (95% CI: 2.73–3.32) with improved heterogeneity from I²=75% (95% CI: 65–83%) to 50% (95% CI: 23–68%) (Supplemental Figure 2). On univariate meta-regression analysis, longer study follow-up time was associated with a lower HCC incidence (all P≤.004, Supplemental Table 1). On multivariate meta-regression analysis, longer study follow-up time and studies including HIV co-infection were associated with a lower HCC incidence, while studies including possible HBV co-infection were associated with a higher HCC incidence (all P≤.03, Supplemental Table 1). The funnel plot (Supplemental Figure 3A) showed slight asymmetry, likely explained by the significant between-study heterogeneity. We did not find evidence of publication bias based on the Egger’s regression test (intercept 0.18, t=0.31, P=.76).

In studies on patients without cirrhosis, no studies were eligible for removal during outlier analysis. Studies including patients with possible HIV co-infection had a statistically significant increase in HCC incidence (P=.01) on univariate meta-regression analysis, but this was no longer significant in multivariate analysis (Supplemental Table 2). The funnel plot (Supplemental Figure 3B) showed slight asymmetry, likely explained by the significant between-study heterogeneity and smaller number of studies included in this analysis. We did not find evidence of publication bias based on the Egger’s regression test (intercept −0.09, t=−0.11, P=.91).

DISCUSSION

In this systematic review and meta-analysis, we summarized HCC incidence rates and clinical outcomes in patients with HCV who achieved SVR after DAA therapy (31 studies: 27,711 patients with cirrhosis; 11 studies: 32,123 patients without cirrhosis), by cirrhosis status and fibrosis stage. HCC incidence was highest among patients with cirrhosis (2.99 per 100 person-years) and much lower in patients with F3 fibrosis (0.63 per 100 person-years) or those with no documented cirrhosis (0.47 per 100 person-years). In patients with F3 fibrosis or no documented cirrhosis, HCC incidence was far below the threshold of >1.32 per 100 person-years considered to result in cost-effective HCC screening in patients with DAA-cured HCV, using current screening modalities¹⁰. In patients with cirrhosis, HCC incidence was highest in studies with <1 year of follow-up (6.17 per 100 person-years) and progressively lower in studies with 1–2 years (2.75 per 100 person-years), 2–3 years (2.90 per 100 person-years), and ≥3 years of follow-up (1.83 per 100 person-years), suggesting that HCC incidence after SVR might be declining as longer time accrues following SVR.

Waziry et al. performed a meta-analysis of 9 DAA studies in 2017 and found an HCC incidence of 2.96 per 100 person-years (95% CI: 1.76–4.96), over a mean study follow-up time of only 1.1 years.⁴⁶ Huang et al. performed a meta-analysis of 16 studies in 2018 and found an HCC incidence of 2.1 per 100 person-years (95% CI: 1.4–3.4), over a median study follow-up time of around 1 year.⁴⁷ The short follow-up time of these prior meta-analyses precluded an assessment of HCC risk over a longer time since SVR (which is more clinically relevant) or an assessment of whether HCC incidence changed as more time accrued following SVR. In our meta-analysis, we included 31 studies with much longer follow-up times (mean 2.1 years), allowing us to better evaluate HCC risk over a longer time period. We included multiple studies with <1 year (4 studies), 1–2 years (18 studies), 2–3 years (4 studies), and ≥3 years (5 studies) of follow-up after SVR. We found that in patients with cirrhosis, HCC risk was significantly higher in studies with <1 year of follow-up (6.17 per 100 person-years) compared to studies with ≥1 year of follow-up (1.83–2.90 per 100 person-years). This trend persisted even after adjusting for geographic region, publication year, presence of an HCC screening program, study quality, study design (prospective vs. retrospective), and HBV or HIV co-infection, in meta-regression analysis. This decline may be partly explained by ascertainment bias related to missed or undiagnosed HCC prior to DAA therapy resulting in a higher rate of HCC within a year of achieving SVR, the possibility of more advanced liver disease in patients included in studies with <1 year of follow-up, or loss to follow-up among patients in the studies with ≥1 year of follow-up. Although all 4 studies with <1 year of follow-up reported baseline HCC screening with abdominal imaging to rule out preexisting HCC before treatment initiation, ultrasound is limited by low sensitivity in detecting early stage HCC (47% sensitivity alone or 63% sensitivity with AFP).⁴⁸ Thus, although our findings suggest that the risk of HCC after DAA-induced SVR may decline over time in patients with cirrhosis, additional studies that provide annual HCC incidence with longer follow-up times will be needed to further evaluate this trend.

In studies of patients without cirrhosis, HCC incidence was 0.47 per 100 person-years (95% CI: 0.32–0.70). Patients with F3 fibrosis, including those identified through non-invasive testing, had a 1.4-fold higher HCC incidence than those with unspecified fibrosis stage, but both rates were lower than the HCC incidence threshold of >1.5 per 100 person-years considered to be cost-effective for HCC screening in cirrhosis patients⁴⁹ or the more recently estimated threshold of >1.32 per 100 person-years specifically for patients with DAA-cured HCV.¹⁰ Although the difference by fibrosis stage was not statistically significant, the pooled HCC incidence rate for F3 fibrosis had an I²=0%, which suggests there was minimal heterogeneity between the included studies and as a result, a greater likelihood of obtaining an accurate pooled estimate. APASL and EASL guidelines currently recommend HCC screening in 6-month intervals for all patients with HCV-related F3 fibrosis.^{9, 50} However, our results suggest that patients without cirrhosis and in particular, those with F3 fibrosis, may not warrant HCC screening, given the low HCC incidence and lack of between-study heterogeneity. Although HCCs may still arise in patients without cirrhosis, we currently do not have strategies (e.g. biomarkers or risk calculators) to accurately identify patients without cirrhosis who may still have a high enough HCC risk to merit ongoing screening. Thus, the reduced benefits of screening should be weighed against the potential costs, risks, and harms of routine HCC screening in this subpopulation.

HCC incidence also varied significantly by geographic region among patients without cirrhosis. Compared to patients in Europe, those in North America, Asia, and the Middle East had a 2.3-fold, 2.6-fold, and 3.5-fold higher HCC incidence, respectively. The higher rates in Asia may be partly explained by the higher prevalence of HBV co-infection in Asian countries⁵¹ which can further increase the risk of developing HCC even in the absence of cirrhosis.⁵² APASL also currently recommends HCC screening in patients with F0-F2 fibrosis, in 6-month intervals for at least 2 years after HCV treatment.⁹ In the Middle East, higher HCC incidence may be related to the higher prevalence of HCV genotype 3 and 4.^53–55 HCV genotype 3 in particular, has been associated with a greater risk of cirrhosis and HCC compared to genotype 1.⁵⁶ The influence of other risk-factors for HCC that vary by region, including other types of liver disease, can also be considered.

Overall, the majority of included studies were of good quality based on the NOS, but the risk of misclassification bias cannot be excluded, as studies varied in their definitions of cirrhosis and F3 fibrosis. Ascertainment bias related to HCC should also be considered, as studies varied in their approach to HCC detection before and after treatment initiation. We did perform an additional analysis of just the 11 studies that included both patients with and without cirrhosis, as these patients would presumably have undergone similar pre- and post-treatment HCC evaluations. Even in this analysis, HCC incidence was much higher at 3.49 per 100 person-years (95% CI: 2.64–4.61; I² =83%, 95% CI: 71–90%) than in patients without cirrhosis at 0.47 per 100 person-years (95% CI: 0.32–0.70), suggesting that detection bias alone is unlikely to explain the difference in HCC incidence by cirrhosis status.

There was also no evidence of publication bias, but our results were limited by substantial heterogeneity. While we did perform multiple sensitivity analyses to evaluate this heterogeneity, we were unable to obtain a homogenous HCC incidence estimate for patients by cirrhosis status. We did find that in patients with cirrhosis, the removal of 3 outlier studies resulted in a similar HCC incidence of 3.01 per 100 person-years, with significant improvement in I² from 75% to 50%. This suggests that at least a part of the heterogeneity may be attributable to just a few of the studies included in this meta-analysis. The heterogeneity may also be related to patient-level and study differences. We found that I² varied during subgroup analyses based on geographic region, study follow-up time, study quality, study design, HCC screening program, and viral co-infection. For example, although studies reporting a post-treatment HCC screening program had lower levels of heterogeneity than those without a program, individual patient adherence to screening could not be confirmed. Similarly, patient-specific follow-up time could not be assessed. This type of heterogeneity may be better addressed with an individual patient data-meta-analysis, which could not be performed as patient level data was not available for all included studies.

Our systematic review and meta-analysis has several strengths. First, we obtained a homogenous pooled estimate for HCC incidence in patients with F3 fibrosis (I²=0%). Second, we provided an updated assessment of HCC incidence rates in patients who achieved DAA-induced SVR, using a larger number of studies with longer follow-up times than what has been described in prior meta-analyses. Third, though limited by heterogeneity, we attempted to provide subpopulation data by summarizing HCC incidence and mortality rates by cirrhosis status. We also performed subgroup analyses by fibrosis status to further delineate HCC incidence by F3 vs. unspecified fibrosis stage. Fourth, we provided data on HCC tumor burden, staging, and treatment modality in patients with cirrhosis diagnosed with new HCC. Finally, we excluded studies that did not provide clear SVR or HCC data by cirrhosis status, to minimize selection bias and maximize the likelihood of obtaining accurate pooled estimates of HCC incidence by cirrhosis status, in individuals who achieved SVR.

Our study did have several limitations. First, we could not gather detailed patient, tumor, or mortality data from all included studies, as many studies did not provide disaggregated patient and clinical data by both cirrhosis and SVR status. As a result, patient-level characteristics could not be considered as covariates in our analysis. Second, risk of misclassification bias remains, as studies varied in how they defined cirrhosis and F3 fibrosis. Notably, all studies defined F3 fibrosis using noninvasive methods, so it is possible that a biopsy may have shown less fibrosis or cirrhosis. Third, despite performing several sensitivity analyses to investigate the substantial heterogeneity found in our meta-analysis, we could not significantly reduce the heterogeneity in the pooled HCC incidence estimates by cirrhosis stage. Fourth, we were unable to further analyze HCC incidence among patients with cirrhosis by Child-Pugh score, which may influence liver-related mortality, access to transplantation, and thereby time at risk for HCC, as this data was not available from most studies. Finally, the number of patients included in the meta-analysis with F3 fibrosis was small compared to those with cirrhosis or unspecified fibrosis stage.

In conclusion, among patients with HCV who achieved SVR after DAA therapy, the incidence of HCC was very high (2.99 per 100 person-years) in patients with cirrhosis. HCC incidence appeared to progressively decline as more time accrued during follow-up. However, these findings were limited by substantial between-study heterogeneity. Future meta-analyses may benefit from pooling individualized patient data to better evaluate HCC incidence rates after HCV eradication in patients with cirrhosis. In patients without cirrhosis or with F3 fibrosis, heterogeneity was minimal, and HCC incidence was far below the cost-effective 1.5% per year threshold for HCC screening. This is especially important in patients with F3 fibrosis, as these patients currently receive screening indefinitely at 6-month intervals in many countries. In patients with F3 fibrosis, the low HCC incidence rate coupled with the lack of between-study heterogeneity provides strong evidence that routine HCC screening may not be warranted; the potential costs, risks, and harms of screening should be weighed against the reduced benefits of screening in this subpopulation.

WHAT YOU NEED TO KNOW.

Background:

Treatment of hepatitis C virus (HCV) infection can reduce the risk of liver cancer. We pooled prior studies to assess the risk of liver cancer after HCV cure, by liver disease severity.

Findings:

Patients with cirrhosis have very high rates of liver cancer even after HCV eradication; this risk may decline slightly as years accrue from the time of HCV cure. Patients without cirrhosis have much lower rates of liver cancer; this risk is far below current thresholds associated with cost-effective liver cancer screening. There was substantial study heterogeneity in this meta-analysis, except in studies of patients with “advanced” F3 fibrosis.

Implications for patient care:

Patients with cirrhosis should continue to undergo liver cancer screening after HCV cure due to their high HCC incidence. In patients without cirrhosis, particularly in those with “advanced” F3 fibrosis, HCC incidence after HCV cure is much lower and below current thresholds believed to result in cost-effective screening.

Abbreviations

AASLD

American Association for the Study of Liver Diseases

APASL

Asian Pacific Association for the Study of the Liver

CI

Confidence interval

CT

Computed tomography

DAA

Direct-acting antivirals

EASL

European Association for the Study of the Liver

HBV

Hepatitis B virus

HCC

Hepatocellular carcinoma

HCV

Hepatitis C virus

MRI

Magnetic resonance imaging

NOS

Newcastle-Ottawa Quality Assessment scale

SVR

Sustained virologic response