Article Text
PDF
Abstract
Background The influence of direct-acting antiviral (DAA) therapy for chronic hepatitis C virus on the risk of hepatocellular carcinoma (HCC) is conflicting.
Methods We conducted a systematic review and meta-analysis to determine the incidence or recurrence of HCC associated with oral DAA therapy. We searched PubMed, Scopus, Embase from inception to August 2017 to identify observational studies reporting on HCC among patients treated with DAAs. Two independent reviewers extracted data and assessed the risk of bias. Data were pooled by random-effects model. The primary outcome was the proportion of participants with incidence or recurrence of HCC (PROSPERO number CRD42017057040).
Results After reviewing 2080 citations, we included 8 controlled studies and 36 uncontrolled studies. The pooled proportion for incident HCC was 1.5 % (95% CI 1.0% to 2.1%; I2=90.1%; n= 542/39 145) from 18 uncontrolled studies and 3.3% (95% CI 1.2% to 9%; I2 =96%; n=109/6909) from 5 controlled studies, respectively. The pooled proportion for recurrent HCC was 16.7% (95% CI 10.2% to 26%; I2=84.8%; n=136/867) from 12 uncontrolled studies and 20.1% (95% CI 5.5% to 52.1%; I2=87.5%; n=36/225) from 3 controlled studies, respectively. There was no statistically significant effect on the risk of recurrent HCC (OR 0.50, 95%CI 0.16 to 1.59; I2 =73.4%) in a meta-analysis of three studies.
Conclusions Our findings show low proportion of incident HCC, but high proportion of recurrent HCC on treatment with DAAs. Continued active surveillance for HCC after treatment with DAAs remains prudent.
- hepatocellular carcinoma
- antiviral therapy
Statistics from Altmetric.com
Significance of this study
What is already known on this topic?
There is conflicting evidence on the association between oral direct-acting antivirals (DAAs) for chronic hepatitis C and the development of hepatocellular carcinoma (HCC).
What this study adds?
Our findings show low proportion of incident HCC, but high proportion of recurrent HCC on treatment with DAAs.
Continued active surveillance for HCC after treatment with DAAs remains prudent.
Background
Approximately 2.7–3.9 million patients suffer from chronic hepatitis C in the USA, a contributing cause for 19 659 deaths in 2014.1 The National Academy has set a goal of eradication of chronic hepatitis C by 2030 as a high priority because it is the leading cause of infectious disease mortality.
The direct-acting antivirals (DAAs) are considered a novel and innovative treatment for chronic active hepatitis C, which affects millions of people worldwide. Several randomised controlled trials (RCTs) have established their efficacy on sustained virologic response (SVR) 12 weeks or 24 weeks after treatment. These agents have excellent tolerability for patients and have the unique ability to achieve viral clearance and maintain SVR remission. While they have the potential to reduce long-term complications of end-stage liver disease, a recent Cochrane review of 157 trials identified the lack of evidence on long-term clinical outcomes.2 RCTs have been of limited duration and are inadequately powered to assess long-term outcomes and complications of chronic active hepatitis such as hepatocellular carcinoma (HCC).
The influence of SVR induced by oral DAAs on the risk of HCC among patients with chronic hepatitis c is conflicting and is mainly derived from small case series and observational studies, and few systematic reviews have rigorously and systematically evaluated this association.3
Objective
Our objective was to determine the incidence or recurrence of HCC associated with oral DAA therapy among patients with chronic active hepatitis C in real-world settings.
Methods
Systematic review registration
We carried out a systematic review and meta-analysis according to a prespecified protocol. The protocol was registered at PROSPERO 2017 CRD42017057040. We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses checklist for reporting the results.
Eligibility criteria
Inclusions: We included any observational study including cohort and case–control studies reporting on the association between DAA use and incidence and recurrence of HCC. The RCTs have been of limited duration and were not designed to measure HCC and were not the focus of this review. A recent Cochrane review did not find sufficient data in the trials to inform on this outcome.2 Exclusions: We excluded case reports, cross-sectional studies, narrative reviews and animal or pharmacokinetic or pharmacodynamic studies. We also excluded non-English articles, and those for which full-text articles cannot be retrieved. For the analysis of incident HCC, we limited the sample size of uncontrolled studies to studies with sample size >100 as smaller cohorts are unlikely to provide reliable estimates of the incidence of HCC, which is a rare outcome. We did not impose such sample size restrictions on the outcome of recurrent HCC.
Participants
We included adult participants with chronic hepatitis C of any genotype treated with any of the DAA agents alone or in combination. We did not impose any restrictions based on Child–Pugh Curtis class and cirrhosis.
Interventions
We examined the DAAs and their combinations shown in our protocol.
Comparators
These comprised of the group of participants not receiving DAAs. We also included studies in which IFN-based regimens were administered in both groups if the difference between the groups was the DAA or DAA combination alone.
Outcomes
The co-primary outcomes were the incidence and recurrence of HCC as defined in the primary studies. We required a minimum of 12 weeks of treatment duration. We did not impose any lag period after exposure as the exact timing of occurrence of HCC after oral DAA exposure is unknown.
Systematic searches
We searched multiple databases, including PubMed, Embase and Scopus, using the search strategy outlined in Supporting Information through 24 August 2017. An outline of the search strategy for PubMed is shown in the protocol. It was modified and adapted for various databases. Details of the search strategy for each database are shown in the Supporting Information, which also identifies the oral DAAs that were included. Additional searches were carried out by manually looking at the reference lists of included studies.
Data extraction
Studies were selected for inclusion based on the review of titles and abstracts that were exported into a web-based data management repository (www.rayyan.qcri.org).4 Two reviewers were independently involved in all stages of study selection, data extraction and quality assessment. All discrepancies were resolved with agreement after rechecking the source papers and further discussion among the reviewers with full consensus prior to inclusion. Data to be extracted included study design, location of the study, drug exposure, mean age of participants, duration of drug use, and duration of follow-up and method of outcome ascertainment.
Risk of bias assessment
We used an adapted version of the Newcastle–Ottawa Scale for evaluating the quality of included studies.5 We also used the Egger’s test and funnel plot to evaluate for publication bias.
Data synthesis
All data were extracted into a prespecified Excel data sheet. The unit of the analysis was the individual participant. We conducted both qualitative and quantitative syntheses. We considered reporting the results of meta-analysis when two or more studies could be pooled in the absence of both clinical heterogeneity and statistical heterogeneity. Examination for clinical heterogeneity included examination for differences in population, intervention, follow-up duration and other characteristics. Statistical heterogeneity was estimated using the I2 statistic. We considered both random-effects and fixed-effects meta-analyses. We conducted a meta-analysis of proportion to determine incidence and recurrence rates which were reported as pooled proportions. The primary analysis was focused on studies of more than 24 weeks follow-up to allow for a sufficient duration of follow-up to detect HCC. In secondary analyses, we pooled studies irrespective of duration of follow-up. For studies with comparators, we conduct meta-analysis using risk ratios (RRs) or OR when appropriate. All analyses were conducted in Comprehensive Meta-Analysis, Englewood, New Jersey and StatsDirect Statistical Software (England, StatsDirect, 2013)
Strength of evidence rating
We also rated the strength of evidence on each of the primary outcomes using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach using the major domains of risk of bias, imprecision, inconsistency, strength of association and publication bias.6
Results
Systematic search
We screened 2080 citations and identified 8 controlled studies and 36 uncontrolled studies that were eligible for inclusion. The process of study selection is shown in figure 1. The baseline characteristics of the uncontrolled and controlled studies are shown in online supplementary table 1 and online supplementary table 2, respectively. The incidence and recurrence of HCC associated with oral DAAs in controlled and uncontrolled studies are shown in online supplementary table 3 and online supplementary table 4, respectively. The risk of bias of all included studies is shown in online supplementary table 5.
Supplementary file 1
Supplementary file 2
Supplementary file 3
Supplementary file 4
Supplementary file 5
Preferred Reporting Items for Systematic Review and Meta-Analysis flow sheet of included studies. HCC, hepatocellular carcinoma; RCTs, randomised controlled trials.
Study characteristics
There were 29 uncontrolled studies that evaluated incident HCC,7–36 while 12 studies reported on recurrent HCC.9–11 14 17 29 37–42 Some studies reported on both incident HCC and recurrent HCC. The uncontrolled studies encompassed a diverse range of patients, with some single-centre studies but also many other multicentre or global cohorts. Treatment durations tended to be 12–24 weeks with different antiviral combinations in the cohorts. Follow-up was also highly varied, ranging from 12 weeks to 70 months. Among the 29 uncontrolled studies, 18 studies had a minimum follow-up duration of 24 weeks. The mean age of participants in each study ranged from about 50 years to 72 years.
Five controlled studies reported on incident HCC in those receiving DAA as compared with controls.22 38 43–45 Recurrent HCC was monitored in three studies, two of which were part of the ANRS cohort (HEPATHER and CirviR).11 46 47 The controlled observational studies were conducted in the UK, Japan, the USA and France. Treatment duration was typically 12–24 weeks, whereas follow-up range was up to 416 weeks. The average age of participants in the studies varied from 51 years to 73 years. A wide variety of oral DAAs were evaluated, whereas the control groups tended to be participants who did not have any treatment, historical controls or received combination of IFN and ribavirin.
Primary analyses
Proportion of incident HCC associated with DAAs in uncontrolled cohorts
A total of 542 cases of HCC were reported in the 39 145 participants on oral DAAs among the 18 uncontrolled studies. The random-effects meta-analysis of 18 studies yielded a pooled proportion of 1.5% (95% CI 1.0% to 2.1%; I2=90.1%) for incident HCC with evidence of substantial statistical heterogeneity among the included studies as shown in figure 2.
Random-effects meta-analysis of proportion of incident hepatocellular carcinoma with oral direct -acting antivirals in uncontrolled studies (primary analysis).
Proportion of incident HCC associated with DAAs in controlled cohorts
Among five controlled cohort studies, the pooled proportion of patients who had incident HCC while on treatment with oral DAAs was 3.3% (95% CI 1.2% to 9%; I2=96%), with evidence of substantial statistical heterogeneity among the included studies as shown in figure 3.
Random-effects meta-analysis of proportion of incident hepatocellular carcinoma with oral direct-acting antivirals in controlled cohorts.
Proportion of recurrent HCC associated with DAAs in uncontrolled cohorts
There were 136 cases of recurrent HCC among 867 participants in 12 uncontrolled studies. The random-effects meta-analysis of 12 studies yielded a pooled proportion of 16.7% (95% CI 10.2% to 26%; I2=84.8) with evidence of substantial statistical heterogeneity among the included studies as shown in figure 4. Estimates of individual studies ranged from as low as 0% (in a study that included only eight participants with pre-existing HCC)41 to as high as 42.6%.39
Random-effects meta-analysis of proportion of recurrent hepatocellular carcinoma with oral direct-acting antivirals in uncontrolled cohorts.
Proportion of recurrent HCC associated with DAAs in controlled cohorts
There were 36 cases of recurrent HCC among 225 participants on oral DAAs in the controlled cohort studies, which yielded a pooled proportion of 20.1% (95% CI 5.5% to 52.1%; I2=87.5%) with evidence of substantial statistical heterogeneity among the included studies as shown in figure 5.
Random-effects meta-analysis of proportion of recurrent hepatocellular carcinoma with oral direct-acting antivirals in controlled cohorts.
Relative risk of incident or recurrent HCC associated with oral DAA in controlled cohorts
We did not conduct a meta-analysis of RRs or ORs comparing the oral DAA and control arms on incident HCC because of varying levels of clinical heterogeneity, follow-up and the choice of comparators.22 38 43–45 Among the five controlled cohort studies that reported on incident HCC, there were 36 HCC recurrences in 225 patients having DAA, as compared with 253 cases of HCC among 4828 control patients (5.2%). Kobayashi et al reported no significant differences in newly diagnosed HCC between the DAA group and the IFN/ribavarin (RBV) group at 3-year follow-up and 5-year follow-up.44 Similarly, Li et al reported a HR of 1.07 (95% CI 0.55 to 2.08) for HCC in the DAA group as compared with the IFN group.22
Among the three studies that reported on recurrent HCC,11 46 47 there were 36 HCC recurrences (16%) in 225 patients having DAA, as compared with 80 recurrences (42.3%) in 189 control patients. The relative risk meta-analysis11 46 47 of these three studies showed an OR of 0.50 (95% CI, 0.16 to 1.59; I2=73.4%) for recurrent HCC, which was not statistically significant as shown in figure 6.
Adjusted random-effects meta-analysis of proportion of recurrent hepatocellular carcinoma with oral direct-acting antivirals in controlled cohorts.
Secondary analysis
Proportion of incident HCC associated with DAAs in uncontrolled cohorts irrespective of follow-up duration
A total of 582 cases of HCC were reported in the 48 473 participants on oral DAAs among the 29 uncontrolled studies. The random -effects meta-analysis of 29 studies yielded a pooled proportion of 1.2% (95% CI, 0.09% to 1.7%; I2=87.5%) for incident HCC with evidence of substantial statistical heterogeneity among the included studies as shown in online supplementary figure 1. The proportion varied from as low as 0.06% to 4.83% due to clinical heterogeneity and variation in methods and duration of follow-up.
Supplementary file 7
![[SP7.jpg]](https://fg.bmj.com/content/flgastro/9/4/262/DC6/embed/inline-supplementary-material-6.jpg?download=true){kind=link}
Sensitivity analysis
The results of the fixed-effects analysis for the above outcomes are shown in online supplementary table 6 and were largely similar to the random-effects analysis.
Supplementary file 6
Risk of bias and GRADE assessment
Most of the studies were considered to have reliable data on the actual use of antiviral therapies (online supplementary table 5). Selection of controls was not always well defined and there were differences in the baseline variables between the DAA-treated participants and the control group. The handling of confounding variables and statistical adjustment was poorly reported. Specific active ascertainment for HCC was performed in 6 controlled cohorts and 10 of the uncontrolled cohorts. The lack of active surveillance may result in an underestimate of the risk. The GRADE of evidence was designated as low for each of these outcomes because of imprecision of summary estimates, substantial heterogeneity and high risk of bias.
Discussion
Our findings show that the estimates for incident HCC range from ≈1. 5% in uncontrolled cohorts to ≈3.3% in controlled cohorts in patients with chronic active hepatitis C. Our meta-analysis shows that the estimates for recurrent HCC range from ≈16.7% in uncontrolled cohorts to ≈20.1% in controlled cohorts. Assuming baseline proportions seen in the uncontrolled cohort studies, approximately 1 among 67 participants who use oral DAAs would develop incident HCC, and approximately 1 among 6participants who are oral DAA users would develop recurrent HCC. The meta-analysis of recurrent HCC from controlled studies showed a ≈50% reduction in the risk of HCC, which was not statistically significant. The overall quality of the evidence was low for each of the outcomes and comparisons. In the absence of high-quality controlled cohorts with long-term follow-up, our findings can neither confirm nor rule out an increase or decrease in the risk of incident or recurrent HCC after oral DAA therapy.
Comparison with other studies
There are some similarities and differences between our findings and other meta-analyses.3 A previous meta-analysis reported a low rate of 2.96/100 person-years (PYs) (95% CI 1.76 to 4.96) for HCC occurrence from 9 studies and a rate of 12.16/100 PYs (95% CI 5 to 29.8) for recurrent HCC from 10 studies.3 Our database of included studies for the risk of HCC associated with oral DAAs was substantially larger (n=44), and we included participants irrespective of the stage of liver disease48 and also assessed HCC either as an adverse event or as a prespecified outcome irrespective of whether participants achieved SVR. In contrast, the other meta-analyses were restricted to participants with chronic hepatitis C virus (HCV) who had undergone curative treatment and excluded patients without cirrhosis or those without clear delineation of HCC outcomes.
Our findings should be interpreted in the context of the historical rates of HCC reported with untreated chronic hepatitis C or the risk of HCC reported with IFN-based treatments, although these populations are not directly comparable, as patients treated with oral DAAs have more advanced liver disease as compared with historical IFN-based controls. Among patients with chronic active hepatitis C, the rate of HCC is around 1–3% per year, and higher among those with cirrhosis. In a Japanese cohort study, the presence of HCV antibody was associated with a more than fourfold increased risk of liver cancer (RR 4.09, 95% CI 1.30 to 12.85) (age range=40–79 years; % cirrhosis) over a 3-year period.49 The other meta-analysis reported incidence rates for HCC occurrence (1.14/100 PYs (95%CI 0.86 to 1.52)) and recurrence (9.21/100 PYs (95%CI 7.18 to 11.81)) associated with IFN-based treatment.3 Another meta-analysis of observational studies reported a substantial (≈ 77%) reduction in the risk of HCC with IFN-based treatments in patients with fibrosis and advanced liver disease compared with untreated controls.50
The precise biological mechanisms by which oral DAA may influence the risk of HCC is unknown.51 Alterations in immunosurveillance that remove the protective effect of the HCV may potentiate the risk of HCC by accelerating the process of regeneration. Studies have shown an increase in the levels of vascular endothelial growth factor after 4 weeks of therapy with oral DAA therapy and alterations in balance between inflammatory and anti-inflammatory processes.52 Additionally, it is possible that the immunostimulatory effects of IFN may have offered some protection in earlier studies. Several studies in our meta-analysis included participants who had undergone non-curative treatment of HCC, where HCC may reflect reactivation of residual foci of untreated HCC. The high proportion of recurrent HCC among patients treated with oral DAAs may reflect the underlying severity of the disease being treated, as participants with advanced liver disease are eligible for treatment with DAAs, where such participants may have been ineligible for treatment with IFN-based regimens. Finally, ascertainment biases where some tumours which were previously undetectable were radiographically identified after treatment with oral DAAs may be another potential explanation of our findings.
Limitations
Our meta-analysis has some limitations which primarily reflect the quality of reported data. The duration of follow-up may have been limited to detect the long-term risk of HCC as most of the included studies were of short duration. In the absence of patient-level data, we did not evaluate time to event and determine the influence of geographic location, the presence of genotype 1, alcohol use, concomitant hepatitis B, serum alpha-fetoprotein levels and stage of cirrhosis on the rate of development of HCC, as meta-regression based on summary data may be prone to ecological biases.53 We cannot rule out the presence of length time bias, because the data did not allow us to distinguish between the development of fast-growing versus slow growing tumours and distinguish between early recurrent HCC which may reflect tumour metastases (<2 years) versus late development of second primary tumours (>2 years).51 The estimates for incident HCC may also reflect recurrent HCC as some studies did not clarify that they reliably excluded HCC at baseline. We excluded studies with <100 participants for incident HCC, as these studies may be susceptible to small study effects. We did not evaluate the differences in rates between individual members of the class of DAAs, although there are no biological reasons to suspect any such differences on the risk of HCC. We did not include non-English studies which may have biased our estimates.
Implications
Our findings have implications for clinical practice and research. Our findings demonstrate that recurrent HCC is common in patients who are taking oral DAAs, and clinicians should therefore remain vigilant and implement regular monitoring even after achieving a SVR. Future studies should be designed with adequate follow-up and latency as HCC occurring immediately after treatment may reflect pre-existing HCC.51
Such studies should use standardised definitions of HCC, conduct periodic surveillance using established radiographic protocols and evaluate the influence of DAAs on clinical outcomes beyond SVR.2 Studies should also evaluate the influence of other variables such as the presence of GT1, alcohol use, concomitant hepatitis B, serum alpha-fetoprotein levels and stage of cirrhosis. Study designs that incorporate the evaluation of the potential therapeutic benefits of oral DAAs earlier during the course of chronic hepatitis C before the development of advanced cirrhosis versus strategies that involve treatment with oral DAAs after the onset of more advanced liver disease may offer additional insight. Finally, the influence of SVR induced by these agents on other important patient-centred outcomes such as hepatic encephalopathy, liver failure, gastrointestinal bleeding, portal hypertension, hepatic encephalopathy, hepatorenal syndrome and mortality should also be evaluated in such studies.
Conclusions
Our findings show low proportion of incident HCC, but high proportion of recurrent HCC on treatment with oral DAAs in patients with chronic active hepatitis C. Regardless of whether high proportion of recurrence of HCC reflects the influence of the drug or the severity of the disease, continued active surveillance for HCC after treatment remains prudent.
References
Footnotes
Contributors SS, AN and YKL contributed to the manuscript by planning the study and reviewing the literature. All the authors collected the data. SS did the meta-analysis. All authors contributed to the assessment and interpretation of data. All authors read, revised and approved the final version of the manuscript.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient consent Not required.
Ethics approval We did not obtain ethical approval since the study was a systematic review of summary data.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement All the data are provided as Supplementary Appendix.