Long-Term Treatment Outcomes of Patients Infected With Hepatitis C Virus: A Systematic Review and Meta-analysis of the Survival Benefit of Achieving a Sustained Virological Response

1Hepatitis C virus (HCV) is a significant public health concern with an estimated 185 million people infected worldwide. HCV progression can lead to the development of liver cirrhosis and hepatocellular carcinoma and results in the deaths of over 700 000 people every year. Combined, viral hepatitis kills more people per year than malaria or tuberculosis but has commanded far less attention and access to care and treatment is limited.

2Traditionally, treatment for HCV has composed of dual-therapy with pegylated-interferon and ribavirin. Dual-therapy is associated with poor sustained virological response (SVR) rates, the surrogate marker for cure defined as undetectable HCV RNA 24 weeks following completion of therapy. A robust treatment pipeline has seen the recent approval of highly efficacious interferon-free regimens with a number of other therapy combinations likely to be approved over the next 2 years. These novel treatment regimens will have the potential to transform the treatment landscape. Promisingly, the high response rate is matched in populations typically considered difficult-to-treat, such as those with advanced fibrosis or coinfection with human immunodeficiency virus (HIV).

3Relative to nonresponders or to those untreated, the attainment of an SVR has repeatedly been associated with improved patient outcomes, irrespective of the path to SVR. These include reduced incidence of liver decompensation, hepatocellular carcinoma, and death. Evidence suggests that an SVR does not only prevent the progression of liver disease but is associated with histologic improvements with some studies even reporting the complete resolution of fibrosis after SVR. Moreover, SVR-achievement has been associated with a reduction in extra-hepatic events and a reduction in mortality independent of liver disease.

4Despite the evidence for improved prognosis with SVR, there are some contradictory data suggesting that SVR-achievement does not provide a significant clinical benefit. A number of studies have shown that the risk of progression is not eliminated with viral eradication, with some patients experiencing decompensation or developing hepatocellular carcinoma despite achieving an SVR. Furthermore, some evidence suggests that the improved prognosis associated with SVR may be diminished in certain patient groups such as those with decompensation or HIV coinfection. There is a need for definitive evidence evaluating the clinical benefit of achieving an SVR in a range of populations, especially given the high cost of interferon-free regimens.

5The aim of this study was to systematically review the current literature concerning the survival benefits of achieving SVR through treatment vs the outcomes in nonresponders and relapsers (non-SVR). All-cause mortality was chosen as the endpoint as it is definitive with clear interpretation. Further, given the extra-hepatic benefits of SVR, all-cause mortality may be clinically more relevant than liver-related mortality.

6We evaluated the mortality rates of patients after treatment for chronic HCV to determine whether, and to what extent, SVR is a prognostic factor for subsequent all-cause mortality.

7Studies for inclusion in the review were identified through an electronic search of 2 biomedical literature databases. The databases PubMed and EMBASE were searched for articles published between 1990 and November 2014 using a sensitive search string with keywords including HCV, SVR, and mortality. No language or geographical restrictions were applied. The search was supplemented by a thorough review of the reference lists of all articles fulfilling eligibility criteria and a search of the proceedings from relevant conferences. Conference proceedings were searched for any relevant articles from 2000 to 2014 and included the American Association for the Study of Liver Diseases, European Association for the Study of the Liver, Asian Pacific Association for the Study of the Liver, Conference on Retroviruses and Opportunistic Infections, and the International AIDS Conference. Two independent authors (B. S. and J. S.) reviewed the process, ensuring the papers met the inclusion criteria and independently extracted the data for review. Any disagreements were resolved by consensus or arbitration by a third reviewer.

8Any retrospective or prospective observational study assessing prognosis of HCV with treatment and any randomized controlled trial assessing the impact of SVR vs non-SVR was eligible for inclusion in the study. Participants had to be adults (>18 years old) chronically infected with HCV of any genotype and were treated with any antiviral regimen for the recommended duration. SVR-achievement was defined as undetectable viremia 24 weeks after completion of antiviral therapy (SVR24); all patients with a detectable viral load at the SVR24 time-point, inclusive of those with an end-of-treatment response, were considered nonresponders and were included in the non-SVR arm. Only trials with a post-therapy follow-up of longer than 1 year were included, and only patients alive at the SVR24 time-point were included in the analyses. Studies were to evaluate all deaths irrespective of cause (all-cause mortality); studies restricted to liver-related mortality were excluded from the current review.

9The eligible articles were stratified into 3 patient populations as follows: (1) General: studies of monoinfected patients at all disease stages; (2) Cirrhotic: studies of monoinfected patients with advanced fibrosis or cirrhosis; (3) HIV/HCV coinfected: all studies of HIV/HCV coinfected patients, regardless of baseline fibrosis status. The following details were extracted from all studies: study location, study type, baseline characteristics, number of patients treated and number achieving SVR, number of deaths in each arm, duration of patient follow-up, and where possible, the hazard ratios (HRs) of mortality. Where data were missing, authors were contacted to retrieve the information; studies with missing follow-up time or other essential raw outcome data were excluded if data were not retrievable. In the case of duplicate studies, the report covering the longest time period with the largest population was used.

10Study quality was evaluated using the Quality in Prognosis Studies (QUIPS) tool, which considers the following 6 domains of bias: participation, attrition, prognostic factor measurement (SVR-attainment), outcome measurement (all-cause mortality), confounding, and analysis and reporting. For each study, each domain was considered as having a high, moderate, or low risk of bias based on a list of prompting study aspects. A bias risk for the analysis domain was only determined in those studies reporting adjusted results.

11For each of the 3 populations, the 5-year mortality rate after treatment was calculated for the SVR and non-SVR arms. The log-transformed incidence rate (IR) and corresponding standard error for each study was calculated using the number of events (deaths) and person-years of follow-up (PYFU). A Poisson distribution was assumed for calculation of the standard error and results were pooled using a random-effects model according to the methods of DerSimonian and Laird. The results were converted to 5-year estimates and presented along with the corresponding 95% confidence interval (CI). A 5-year horizon was deemed most appropriate as the follow-up period in the majority of studies did not exceed this time-point (median follow-up 5.4 years [interquartile range {IQR}, 4.9–7.5]). Plots of IR against follow-up time were visually inspected to test the assumption that the mortality rate was constant over this timespan.

12A comparison of the risk of death in the SVR group vs the non-SVR group was conducted by pooling the HRs for mortality. The HRs reported in each study were calculated using Cox proportional hazards models, and both the unadjusted and adjusted HRs were extracted along with the corresponding variances. As above, pooled estimates for the adjusted HRs were computed using a random-effects model. Where necessary, variance was calculated according to the methods of Parmar et al. Heterogeneity across studies was quantitatively assessed using the I2 statistic in accordance with the Cochrane Handbook. All analyses were conducted using Review Manager (RevMan version 5.3; Cochrane Collaboration) and Stata (STATA 12; StataCorp LP).

13The existence of publication bias was assessed using funnel plots. Statistical tests for asymmetry are low powered, and as such, given the small number of studies anticipated per group, funnel plots were interpreted by visual inspection.

14The search strategy initially yielded 4877 articles, of which 4746 were found to be irrelevant and were excluded. A further 11 potential studies were identified through the reference list review and the search of conference proceedings. Of the final 142 articles, 31 (n = 33 360) fitted the criteria for inclusion. The main reasons for exclusion included absence of mortality data, unclear recording of essential outcomes, including follow-up time, number with SVR, and number of deaths, and duplication of studies. Of the final 31 studies, 17 were in patients at any stage of liver fibrosis (general studies; n = 28 398), 9 were in cirrhotic patients (n = 2604), and the remaining 5 studies were of HIV/HCV coinfected patients (n = 2358). The median of the median follow-up time was 5.2 years (IQR, 4.3–7.8) in the general studies, 6.8 years (IQR, 5.8–7.9) in the cirrhotic studies, and 5.0 years (IQR, 4.6–5.2) in the coinfected studies. The majority of studies were carried out in European, Asian, or North American settings. Participants were predominantly male, infected with HCV genotype 1, and between the ages of 40 and 50 at baseline. All participants were treated with interferon or pegylated-interferon, either as monotherapy or in combination with ribavirin. Study characteristics are shown in Table 1.

15Of the 31 included studies, 5.7% of the domains, that is, inclusion, attrition, prognostic factor measurement, outcome measurement, confounding, and analysis and reporting as assessed with the QUIPS tool, showed a high risk of bias, 26.1% showed a moderate risk, and 68.2% showed a low risk of bias (Supplementary Appendix 1). Twenty-three studies showed a moderate-to-high risk of bias in 1 or 2 domains; 6 showed a moderate-to-high risk of bias in 3 or 4 domains. Risks of bias were highest in the domain of prognostic factor measurement (high in 8/31 [25.8%] and moderate in 14/31 [45.2%]), due to follow-up not originating at the SVR time-point. In these studies, follow-up was often measured from initiation of treatment, and in some cases from biopsy that was conducted up to 1 year prior to treatment.

16In the general population, 502 of 12 140 (54 651 PYFU) patients achieving an SVR died during follow-up equating to a pooled IR of 0.4/100PY (95% CI,.2–.7). In comparison, 1708 out of 16 258 (77 130 PYFU) non-SVR patients died (IR = 1.6/100PY, 95% CI, 1.2–2.3).

17In the cirrhotic studies 45 of 778 (5352 PYFU) SVR patients died during follow-up (IR = 1.0/100PY, 95% CI,.7–1.5) vs 404 of 2108 (15 836 PYFU) non-SVR patients (IR = 3.4/100PY, 95% CI, 2.4–4.8). Finally, in the HIV coinfected population 11 of 857 (4333 PYFU) SVR patients (IR = 0.3/100PY, 95% CI,.1–.6) and 161 of 1501 (7683 PYFU) non-SVR patients died during follow-up (IR = 2.4/100PY, 95% CI, 1.3–4.2). Visual observation of the plots of IR against follow-up time showed no association between the length of follow-up and the risk of mortality in either the SVR or non-SVR groups in all 3 populations; it was thus deemed appropriate to determine the 5-year mortality rates from these data.

18As shown in Figure 1, the estimated 5-year mortality rate was significantly lower for patients achieving SVR compared with nonresponders for all 3 patient populations. The difference in mortality rate between SVR and non-SVR was most pronounced in the cirrhotic and coinfected populations.

19Of the 31 studies included, 21 reported HRs for mortality adjusted for potential covariates that may have had an impact on the results. As shown in Table 2, the endpoint analyzed differed between studies. The majority of studies analyzed the rate of all-cause mortality, either alone (n = 12) or including liver-transplantation as a surrogate for mortality (n = 3). Of the remaining 6 studies, 5 evaluated liver-related deaths, and the last study evaluated non-liver related deaths. Furthermore, a number of studies compared mortality risk after SVR with the risk in untreated patients, in contrast with non-SVR (n = 7, all general studies). Most studies conducted a comprehensive analysis, adjusting for a variety of factors that may have impacted results, including age, gender, fibrosis stage, genotype, alcohol use, and comorbidities (Table 2).

20The results of the pooled HR analysis are shown in Figure 2A–C. In all studies SVR-attainment remained a significant predictor of reduced mortality after adjustment for covariates. SVR had the largest protective effect in the coinfected population (HR = 0.21, 95% CI,.10–.45, median follow-up 5.2 years), followed by the cirrhotic population (HR = 0.26, 95% CI,.18–.37, median follow-up 6.8 years), and the general population (HR = 0.33, 95% CI,.23–.46, median follow-up 5.0 years). In the general population considerable heterogeneity between studies was observed (I2 = 76%, P <.0001). As such a subgroup analysis was conducted and it was found that the HR significantly differed when the reference group was an untreated population (HR = 0.19, 95% CI,.13–.28) compared with non-SVR (HR = 0.50, 95% CI,.37–.67; P <.0001). This result was confirmed by the funnel plot analysis which showed 2 distinct subgroups of studies (Supplementary Appendix 2). There was no evidence of heterogeneity between studies in both the cirrhotic and coinfected populations (I2 = 0%), and all studies in these groups compared SVR with non-SVR. Furthermore, based on a funnel plot examination of the cirrhotic and coinfected populations there was no evidence of bias; however, this result should be interpreted with caution due to the small number of studies.

21The results of this large meta-analysis investigating the risk of mortality after treatment for chronic HCV indicate that achieving an SVR significantly reduces the risk of death compared with unsuccessful therapy in a variety of populations. After adjustment for potential confounding factors, an SVR was associated with approximately a 50%, 74%, and 79% decreased risk of all-cause mortality compared with not achieving an SVR in the general, cirrhotic, and coinfected populations respectively. The decrease in risk gives rise to a substantially lower 5-year mortality rate in patients achieving SVR compared with nonresponders. This difference was most pronounced in the cirrhotic and coinfected cohorts. Cumulatively, this evidence suggests that there is a significant survival benefit of attaining an SVR, even in patients with cirrhosis and those coinfected with HIV.

22Interestingly, the 5-year mortality rate was lowest in patients coinfected with HIV achieving an SVR (1.5%), contradicting existing hypotheses that coinfected patients suffer from higher overall mortality than monoinfected patients. This is likely due to the small number of studies evaluating this population, meaning that differences in absolute reductions in risk are more prominent. Indeed, the risk reduction of death is highest in this population, corroborating evidence that attainment of an SVR can prevent the increased rate of liver-complications associated with HIV coinfection.

23All-cause mortality was deemed the most appropriate endpoint for a number of reasons. Firstly, there are a number of extra-hepatic complications of chronic HCV that can result in mortality unrelated to liver events. These manifestations of HCV include Type II diabetes mellitus, rheumatic disorders, and cardiac disease. Mortality associated with extra-hepatic disorders may account for why the mortality estimates in the present study are greater than those previously reported. Second, the use of survival as an endpoint is applicable to both high income countries, and low and middle income countries. The aversion of the need for a liver transplant has been used to justify high prices of treatment for HCV; however, for most people infected with HCV, transplantation is not an option.

24There are a number of limitations to the current analysis. Above all, there is a concern that the group of patients achieving an SVR systematically differ from patients not achieving an SVR in their baseline characteristics, which may in turn affect outcomes. Patients achieving an SVR tend to be younger, with less severe progression of HCV, and with lower comorbidities, characteristics that could result in lower mortality, regardless of SVR. These potential biases were taken in to consideration by presenting adjusted results, which demonstrate a lower risk of mortality after SVR, independent of other factors. There is some uncertainty over the reliability of these results, as due to differences in the data reported in the literature, the estimates combine different endpoints. Additionally, multivariate analysis may not have been adequate, or in studies where extensive multivariate analyses was carried out, the possibility remains that survival benefit is influenced by additional confounding factors. This criticism would likely be exacerbated when comparing patients achieving SVR with those not treated, given that the present comparator, the non-SVR group, were healthy enough to attempt treatment. The most rigorous way to assess the impact of attaining an SVR on mortality would be to conduct a randomized controlled trial comparing treatment with no treatment. This, however is inappropriate given the related ethical concerns. Furthermore, there was a high risk of bias in relation to the origin of follow-up. A number of studies measured follow-up from treatment initiation, or even earlier than this, rather than the SVR time-point, allowing the accruement of PYFU before SVR-attainment. The impact of this would likely be diminished in the pooled HR analysis given that the origin of follow-up was the same for both arms in each individual study.

25The results presented in this analysis are for a 5-year follow-up period due to this being the average follow-up duration. Estimates for a longer timespan would require a greater number of assumptions regarding the relative outcomes between the SVR and non-SVR groups and was thus deemed inappropriate. There is a need for longer-term follow-up to see whether the survival benefit is sustained. Lastly, the current findings are from studies of patients treated with interferon-based treatment, with long-term outcome data currently unavailable for people treated with the more efficacious all-oral therapies.

26The results of this meta-analysis suggest that there is a significant survival benefit of achieving an SVR compared with unsuccessful treatment. Moreover, this benefit is held in patients with cirrhosis and those coinfected with HIV. There are no data to support the notion that the value of achieving SVR is influenced by the means used to achieve it. Although the expectation is that patients achieving SVR with interferon free treatment will have at least as much benefit from SVR as seen in historical studies, post-SVR patients cohorts do not yet have sufficient follow-up time to be helpful. Monitoring these outcomes has been built in to a number of registration trial programs, and further data collection over coming years will be important to build on the studies analyzed here.

27Supplementary materials are available at Clinical Infectious Diseases online (http://cid.oxfordjournals.org). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.